data: sync verified prices 2026-07-27

meta.json

@@ -1,10 +1,10 @@
 {
   "dataset_name": "AICostBudget AI API Pricing Dataset",
   "dataset_version": "1.0.0",
-  "generated_at": "2026-07-09T18:36:17Z",
+  "generated_at": "2026-07-26T19:14:17Z",
   "homepage": "https://aicostbudget.com",
-  "last_verified_at": "2026-07-10T00:00:00Z",
+  "last_verified_at": "2026-07-27T00:00:00Z",
   "model_count": 21,
-  "official_source_count": 9,
+  "official_source_count": 10,
   "provider_count": 7
 }

prices.csv

@@ -1,22 +1,22 @@
 provider_id,model_id,display_name,model_family,status,currency,unit,input,output,cached_input,cache_write,batch_input,batch_output,official_source_url,accessed_at,last_verified_at,effective_from,notes
-anthropic,claude-haiku-4.5,Claude Haiku 4.5,Claude Haiku,active,USD,1M tokens,1.0,5.0,0.1,1.25,0.5,2.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,cache_write stores the 5-minute write rate.
-anthropic,claude-opus-4.8,Claude Opus 4.8,Claude Opus,active,USD,1M tokens,5.0,25.0,0.5,6.25,2.5,12.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,"Anthropic table columns are input, 5-minute cache write, 1-hour cache write, cache read, and output. cache_write stores the 5-minute write rate."
-anthropic,claude-sonnet-4.6,Claude Sonnet 4.6,Claude Sonnet,active,USD,1M tokens,3.0,15.0,0.3,3.75,1.5,7.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,cache_write stores the 5-minute write rate.
-anthropic,claude-sonnet-5-intro,Claude Sonnet 5 introductory pricing,Claude Sonnet,active,USD,1M tokens,2.0,10.0,0.2,2.5,1.0,5.0,https://docs.anthropic.com/en/docs/about-claude/pricing#claude-sonnet-5-introductory-pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,2026-07-05,Introductory pricing is listed by Anthropic as effective through 2026-08-31; standard pricing starts 2026-09-01.
-cohere,aya-expanse-32b,Aya Expanse 32B,Aya,active,USD,1M tokens,0.5,1.5,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Cohere official pricing page states Aya Expanse 8B and 32B API pricing is $0.50/1M input and $1.50/1M output.
-cohere,command-a-plus,Command A+,Command,active,USD,1M tokens,,,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Cohere official pricing page lists Command A+ but the extracted public page did not expose a confirmed token price. Values are null rather than guessed.
-cohere,command-r-plus-08-2024,Command R+ 08-2024,Command R,active,USD,1M tokens,2.5,10.0,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,2024-08-01,Cohere official pricing page lists Command R+ 08-2024 at $2.50/1M input tokens and $10.00/1M output tokens.
-deepseek,deepseek-v4-flash,DeepSeek V4 Flash,DeepSeek V4,active,USD,1M tokens,0.14,0.28,0.0028,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
-deepseek,deepseek-v4-pro,DeepSeek V4 Pro,DeepSeek V4,active,USD,1M tokens,0.435,0.87,0.003625,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
-google-gemini,gemini-2.5-flash,Gemini 2.5 Flash,Gemini 2.5,active,USD,1M tokens,0.3,2.5,0.03,,0.15,1.25,https://ai.google.dev/gemini-api/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Rates are for text/image/video input; Google lists separate audio rates in the official pricing table.
-google-gemini,gemini-2.5-pro,Gemini 2.5 Pro,Gemini 2.5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://ai.google.dev/gemini-api/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Rates are for prompts up to 200k tokens. Google lists higher rates above 200k tokens in the official pricing table.
-mistral-ai,mistral-large,Mistral Large,Mistral Large,active,USD,1M tokens,2.0,6.0,,,1.0,3.0,https://mistral.ai/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Mistral official pricing FAQ states Mistral Large costs $2/M input and $6/M output and batch processing gets a 50% discount.
-openai,gpt-4.1,GPT-4.1,GPT-4.1,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5,GPT-5,GPT-5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5.4-mini,GPT-5.4 mini,GPT-5.4,active,USD,1M tokens,0.75,4.5,0.075,,0.375,2.25,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5.5,GPT-5.5,GPT-5.5,active,USD,1M tokens,5.0,30.0,0.5,,2.5,15.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens. Batch prices are 50% of the listed standard price where shown.
-openai,gpt-5.6-luna,GPT-5.6 Luna,GPT-5.6,active,USD,1M tokens,1.0,6.0,0.1,1.25,0.5,3.0,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,gpt-5.6-sol,GPT-5.6 Sol,GPT-5.6,active,USD,1M tokens,5.0,30.0,0.5,6.25,2.5,15.0,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,gpt-5.6-terra,GPT-5.6 Terra,GPT-5.6,active,USD,1M tokens,2.5,15.0,0.25,3.125,1.25,7.5,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,o3,o3,o-series,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-xai,grok-4.3,Grok 4.3,Grok 4,active,USD,1M tokens,1.25,2.5,,,,,https://docs.x.ai/developers/models,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,xAI official model page lists Grok 4.3 input and output rates per 1M tokens. Cached and batch rates were not shown in the extracted official model card.
+anthropic,claude-haiku-4.5,Claude Haiku 4.5,Claude Haiku,active,USD,1M tokens,1.0,5.0,0.1,1.25,0.5,2.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,cache_write stores the 5-minute write rate.
+anthropic,claude-opus-4.8,Claude Opus 4.8,Claude Opus,active,USD,1M tokens,5.0,25.0,0.5,6.25,2.5,12.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"Anthropic table columns are input, 5-minute cache write, 1-hour cache write, cache read, and output. cache_write stores the 5-minute write rate."
+anthropic,claude-sonnet-4.6,Claude Sonnet 4.6,Claude Sonnet,active,USD,1M tokens,3.0,15.0,0.3,3.75,1.5,7.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,cache_write stores the 5-minute write rate.
+anthropic,claude-sonnet-5-intro,Claude Sonnet 5 introductory pricing,Claude Sonnet,active,USD,1M tokens,2.0,10.0,0.2,2.5,1.0,5.0,https://www.anthropic.com/news/claude-sonnet-5,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,2026-06-30,Introductory pricing applies through 2026-08-31 and standard pricing starts 2026-09-01. V1 must not switch early to the standard $3/M input and $15/M output rates.
+cohere,aya-expanse-32b,Aya Expanse 32B,Aya,active,USD,1M tokens,0.5,1.5,,,,,https://cohere.com/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Cohere official pricing page states Aya Expanse 8B and 32B API pricing is $0.50/1M input and $1.50/1M output.
+cohere,command-a-plus,Command A+,Command,active,USD,1M tokens,,,,,,,https://docs.cohere.com/docs/command-a-plus,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"Cohere's official Command A+ page states that access is free until applicable rate limits are reached and that production use is available through Model Vault. Another official Cohere page shows per-token pricing for the same model ID, so V1 keeps token pricing null until the official documentation is reconciled."
+cohere,command-r-plus-08-2024,Command R+ 08-2024,Command R,active,USD,1M tokens,2.5,10.0,,,,,https://cohere.com/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,2024-08-01,Cohere official pricing page lists Command R+ 08-2024 at $2.50/1M input tokens and $10.00/1M output tokens.
+deepseek,deepseek-v4-flash,DeepSeek V4 Flash,DeepSeek V4,active,USD,1M tokens,0.14,0.28,0.0028,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
+deepseek,deepseek-v4-pro,DeepSeek V4 Pro,DeepSeek V4,active,USD,1M tokens,0.435,0.87,0.003625,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
+google-gemini,gemini-2.5-flash,Gemini 2.5 Flash,Gemini 2.5,active,USD,1M tokens,0.3,2.5,0.03,,0.15,1.25,https://ai.google.dev/gemini-api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Rates are for text/image/video input; Google lists separate audio rates in the official pricing table.
+google-gemini,gemini-2.5-pro,Gemini 2.5 Pro,Gemini 2.5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://ai.google.dev/gemini-api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Rates are for prompts up to 200k tokens. Google lists higher rates above 200k tokens in the official pricing table.
+mistral-ai,mistral-large,Mistral Large 3,Mistral Large,active,USD,1M tokens,0.5,1.5,0.05,,0.25,0.75,https://mistral.ai/pricing/api/,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,mistral-large-latest currently maps to Mistral Large 3. Standard price is $0.50/M input and $1.50/M output. Cached input receives a 90% discount. Batch processing receives a 50% discount.
+openai,gpt-4.1,GPT-4.1,GPT-4.1,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5,GPT-5,GPT-5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5.4-mini,GPT-5.4 mini,GPT-5.4,active,USD,1M tokens,0.75,4.5,0.075,,0.375,2.25,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5.5,GPT-5.5,GPT-5.5,active,USD,1M tokens,5.0,30.0,0.5,,2.5,15.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens. Batch prices are 50% of the listed standard price where shown.
+openai,gpt-5.6-luna,GPT-5.6 Luna,GPT-5.6,active,USD,1M tokens,1.0,6.0,0.1,1.25,0.5,3.0,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,gpt-5.6-sol,GPT-5.6 Sol,GPT-5.6,active,USD,1M tokens,5.0,30.0,0.5,6.25,2.5,15.0,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,gpt-5.6-terra,GPT-5.6 Terra,GPT-5.6,active,USD,1M tokens,2.5,15.0,0.25,3.125,1.25,7.5,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,o3,o3,o-series,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+xai,grok-4.3,Grok 4.3,Grok 4,active,USD,1M tokens,1.25,2.5,0.2,,1.0,2.0,https://docs.x.ai/developers/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,V1 stores short-context pricing. The long-context threshold is 200K tokens. Grok 4.3 batch pricing receives a 20% discount. Long-context prices are not stored in the current short-context fields.

prices.json

@@ -2,9 +2,9 @@
   "dataset_name": "AICostBudget AI API Pricing Dataset",
   "dataset_version": "1.0.0",
   "description": "Open, machine-readable pricing data for LLM and AI APIs.",
-  "generated_at": "2026-07-09T18:36:17Z",
+  "generated_at": "2026-07-26T19:14:17Z",
   "homepage": "https://aicostbudget.com",
-  "last_verified_at": "2026-07-10T00:00:00Z",
+  "last_verified_at": "2026-07-27T00:00:00Z",
   "licenses": {
     "code": "MIT",
     "data": "CC BY 4.0"
@@ -12,10 +12,10 @@
   "model_count": 21,
   "models": [
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Claude Haiku 4.5",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Claude Haiku",
       "model_id": "claude-haiku-4.5",
       "notes": "cache_write stores the 5-minute write rate.",
@@ -34,10 +34,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Claude Opus 4.8",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Claude Opus",
       "model_id": "claude-opus-4.8",
       "notes": "Anthropic table columns are input, 5-minute cache write, 1-hour cache write, cache read, and output. cache_write stores the 5-minute write rate.",
@@ -56,10 +56,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Claude Sonnet 4.6",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Claude Sonnet",
       "model_id": "claude-sonnet-4.6",
       "notes": "cache_write stores the 5-minute write rate.",
@@ -78,14 +78,14 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Claude Sonnet 5 introductory pricing",
-      "effective_from": "2026-07-05",
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "effective_from": "2026-06-30",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Claude Sonnet",
       "model_id": "claude-sonnet-5-intro",
-      "notes": "Introductory pricing is listed by Anthropic as effective through 2026-08-31; standard pricing starts 2026-09-01.",
-      "official_source_url": "https://docs.anthropic.com/en/docs/about-claude/pricing#claude-sonnet-5-introductory-pricing",
+      "notes": "Introductory pricing applies through 2026-08-31 and standard pricing starts 2026-09-01. V1 must not switch early to the standard $3/M input and $15/M output rates.",
+      "official_source_url": "https://www.anthropic.com/news/claude-sonnet-5",
       "pricing": {
         "batch_input": 1.0,
         "batch_output": 5.0,
@@ -100,10 +100,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Aya Expanse 32B",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Aya",
       "model_id": "aya-expanse-32b",
       "notes": "Cohere official pricing page states Aya Expanse 8B and 32B API pricing is $0.50/1M input and $1.50/1M output.",
@@ -122,14 +122,14 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Command A+",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Command",
       "model_id": "command-a-plus",
-      "notes": "Cohere official pricing page lists Command A+ but the extracted public page did not expose a confirmed token price. Values are null rather than guessed.",
-      "official_source_url": "https://cohere.com/pricing",
+      "notes": "Cohere's official Command A+ page states that access is free until applicable rate limits are reached and that production use is available through Model Vault. Another official Cohere page shows per-token pricing for the same model ID, so V1 keeps token pricing null until the official documentation is reconciled.",
+      "official_source_url": "https://docs.cohere.com/docs/command-a-plus",
       "pricing": {
         "batch_input": null,
         "batch_output": null,
@@ -144,10 +144,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Command R+ 08-2024",
       "effective_from": "2024-08-01",
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Command R",
       "model_id": "command-r-plus-08-2024",
       "notes": "Cohere official pricing page lists Command R+ 08-2024 at $2.50/1M input tokens and $10.00/1M output tokens.",
@@ -166,10 +166,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "DeepSeek V4 Flash",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "DeepSeek V4",
       "model_id": "deepseek-v4-flash",
       "notes": "DeepSeek labels input cache miss as input price and cache hit as cached_input.",
@@ -188,10 +188,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "DeepSeek V4 Pro",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "DeepSeek V4",
       "model_id": "deepseek-v4-pro",
       "notes": "DeepSeek labels input cache miss as input price and cache hit as cached_input.",
@@ -210,10 +210,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Gemini 2.5 Flash",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Gemini 2.5",
       "model_id": "gemini-2.5-flash",
       "notes": "Rates are for text/image/video input; Google lists separate audio rates in the official pricing table.",
@@ -232,10 +232,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Gemini 2.5 Pro",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Gemini 2.5",
       "model_id": "gemini-2.5-pro",
       "notes": "Rates are for prompts up to 200k tokens. Google lists higher rates above 200k tokens in the official pricing table.",
@@ -254,32 +254,32 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
-      "display_name": "Mistral Large",
+      "accessed_at": "2026-07-27T00:00:00Z",
+      "display_name": "Mistral Large 3",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Mistral Large",
       "model_id": "mistral-large",
-      "notes": "Mistral official pricing FAQ states Mistral Large costs $2/M input and $6/M output and batch processing gets a 50% discount.",
-      "official_source_url": "https://mistral.ai/pricing",
+      "notes": "mistral-large-latest currently maps to Mistral Large 3. Standard price is $0.50/M input and $1.50/M output. Cached input receives a 90% discount. Batch processing receives a 50% discount.",
+      "official_source_url": "https://mistral.ai/pricing/api/",
       "pricing": {
-        "batch_input": 1.0,
-        "batch_output": 3.0,
+        "batch_input": 0.25,
+        "batch_output": 0.75,
         "cache_write": null,
-        "cached_input": null,
+        "cached_input": 0.05,
         "currency": "USD",
-        "input": 2.0,
-        "output": 6.0,
+        "input": 0.5,
+        "output": 1.5,
         "unit": "1M tokens"
       },
       "provider_id": "mistral-ai",
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-4.1",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-4.1",
       "model_id": "gpt-4.1",
       "notes": "OpenAI official pricing page lists standard and batch rates per 1M tokens.",
@@ -298,10 +298,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5",
       "model_id": "gpt-5",
       "notes": "OpenAI official pricing page lists standard and batch rates per 1M tokens.",
@@ -320,10 +320,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5.4 mini",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5.4",
       "model_id": "gpt-5.4-mini",
       "notes": "OpenAI official pricing page lists standard and batch rates per 1M tokens.",
@@ -342,10 +342,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5.5",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5.5",
       "model_id": "gpt-5.5",
       "notes": "OpenAI official pricing page lists standard and batch rates per 1M tokens. Batch prices are 50% of the listed standard price where shown.",
@@ -364,10 +364,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-10T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5.6 Luna",
       "effective_from": null,
-      "last_verified_at": "2026-07-10T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5.6",
       "model_id": "gpt-5.6-luna",
       "notes": "OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices.",
@@ -386,10 +386,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-10T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5.6 Sol",
       "effective_from": null,
-      "last_verified_at": "2026-07-10T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5.6",
       "model_id": "gpt-5.6-sol",
       "notes": "OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices.",
@@ -408,10 +408,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-10T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "GPT-5.6 Terra",
       "effective_from": null,
-      "last_verified_at": "2026-07-10T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "GPT-5.6",
       "model_id": "gpt-5.6-terra",
       "notes": "OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices.",
@@ -430,10 +430,10 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "o3",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "o-series",
       "model_id": "o3",
       "notes": "OpenAI official pricing page lists standard and batch rates per 1M tokens.",
@@ -452,19 +452,19 @@
       "status": "active"
     },
     {
-      "accessed_at": "2026-07-05T00:00:00Z",
+      "accessed_at": "2026-07-27T00:00:00Z",
       "display_name": "Grok 4.3",
       "effective_from": null,
-      "last_verified_at": "2026-07-05T00:00:00Z",
+      "last_verified_at": "2026-07-27T00:00:00Z",
       "model_family": "Grok 4",
       "model_id": "grok-4.3",
-      "notes": "xAI official model page lists Grok 4.3 input and output rates per 1M tokens. Cached and batch rates were not shown in the extracted official model card.",
-      "official_source_url": "https://docs.x.ai/developers/models",
+      "notes": "V1 stores short-context pricing. The long-context threshold is 200K tokens. Grok 4.3 batch pricing receives a 20% discount. Long-context prices are not stored in the current short-context fields.",
+      "official_source_url": "https://docs.x.ai/developers/pricing",
       "pricing": {
-        "batch_input": null,
-        "batch_output": null,
+        "batch_input": 1.0,
+        "batch_output": 2.0,
         "cache_write": null,
-        "cached_input": null,
+        "cached_input": 0.2,
         "currency": "USD",
         "input": 1.25,
         "output": 2.5,
@@ -474,17 +474,18 @@
       "status": "active"
     }
   ],
-  "official_source_count": 9,
+  "official_source_count": 10,
   "official_sources": [
     "https://ai.google.dev/gemini-api/docs/pricing",
     "https://api-docs.deepseek.com/quick_start/pricing",
     "https://cohere.com/pricing",
     "https://developers.openai.com/api/docs/pricing",
     "https://docs.anthropic.com/en/docs/about-claude/pricing",
-    "https://docs.anthropic.com/en/docs/about-claude/pricing#claude-sonnet-5-introductory-pricing",
-    "https://docs.x.ai/developers/models",
-    "https://mistral.ai/pricing",
-    "https://platform.openai.com/docs/pricing"
+    "https://docs.cohere.com/docs/command-a-plus",
+    "https://docs.x.ai/developers/pricing",
+    "https://mistral.ai/pricing/api/",
+    "https://platform.openai.com/docs/pricing",
+    "https://www.anthropic.com/news/claude-sonnet-5"
   ],
   "provider_count": 7,
   "providers": [

train.csv

@@ -1,22 +1,22 @@
 provider_id,model_id,display_name,model_family,status,currency,unit,input,output,cached_input,cache_write,batch_input,batch_output,official_source_url,accessed_at,last_verified_at,effective_from,notes
-anthropic,claude-haiku-4.5,Claude Haiku 4.5,Claude Haiku,active,USD,1M tokens,1.0,5.0,0.1,1.25,0.5,2.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,cache_write stores the 5-minute write rate.
-anthropic,claude-opus-4.8,Claude Opus 4.8,Claude Opus,active,USD,1M tokens,5.0,25.0,0.5,6.25,2.5,12.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,"Anthropic table columns are input, 5-minute cache write, 1-hour cache write, cache read, and output. cache_write stores the 5-minute write rate."
-anthropic,claude-sonnet-4.6,Claude Sonnet 4.6,Claude Sonnet,active,USD,1M tokens,3.0,15.0,0.3,3.75,1.5,7.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,cache_write stores the 5-minute write rate.
-anthropic,claude-sonnet-5-intro,Claude Sonnet 5 introductory pricing,Claude Sonnet,active,USD,1M tokens,2.0,10.0,0.2,2.5,1.0,5.0,https://docs.anthropic.com/en/docs/about-claude/pricing#claude-sonnet-5-introductory-pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,2026-07-05,Introductory pricing is listed by Anthropic as effective through 2026-08-31; standard pricing starts 2026-09-01.
-cohere,aya-expanse-32b,Aya Expanse 32B,Aya,active,USD,1M tokens,0.5,1.5,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Cohere official pricing page states Aya Expanse 8B and 32B API pricing is $0.50/1M input and $1.50/1M output.
-cohere,command-a-plus,Command A+,Command,active,USD,1M tokens,,,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Cohere official pricing page lists Command A+ but the extracted public page did not expose a confirmed token price. Values are null rather than guessed.
-cohere,command-r-plus-08-2024,Command R+ 08-2024,Command R,active,USD,1M tokens,2.5,10.0,,,,,https://cohere.com/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,2024-08-01,Cohere official pricing page lists Command R+ 08-2024 at $2.50/1M input tokens and $10.00/1M output tokens.
-deepseek,deepseek-v4-flash,DeepSeek V4 Flash,DeepSeek V4,active,USD,1M tokens,0.14,0.28,0.0028,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
-deepseek,deepseek-v4-pro,DeepSeek V4 Pro,DeepSeek V4,active,USD,1M tokens,0.435,0.87,0.003625,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
-google-gemini,gemini-2.5-flash,Gemini 2.5 Flash,Gemini 2.5,active,USD,1M tokens,0.3,2.5,0.03,,0.15,1.25,https://ai.google.dev/gemini-api/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Rates are for text/image/video input; Google lists separate audio rates in the official pricing table.
-google-gemini,gemini-2.5-pro,Gemini 2.5 Pro,Gemini 2.5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://ai.google.dev/gemini-api/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Rates are for prompts up to 200k tokens. Google lists higher rates above 200k tokens in the official pricing table.
-mistral-ai,mistral-large,Mistral Large,Mistral Large,active,USD,1M tokens,2.0,6.0,,,1.0,3.0,https://mistral.ai/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,Mistral official pricing FAQ states Mistral Large costs $2/M input and $6/M output and batch processing gets a 50% discount.
-openai,gpt-4.1,GPT-4.1,GPT-4.1,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5,GPT-5,GPT-5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5.4-mini,GPT-5.4 mini,GPT-5.4,active,USD,1M tokens,0.75,4.5,0.075,,0.375,2.25,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-openai,gpt-5.5,GPT-5.5,GPT-5.5,active,USD,1M tokens,5.0,30.0,0.5,,2.5,15.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens. Batch prices are 50% of the listed standard price where shown.
-openai,gpt-5.6-luna,GPT-5.6 Luna,GPT-5.6,active,USD,1M tokens,1.0,6.0,0.1,1.25,0.5,3.0,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,gpt-5.6-sol,GPT-5.6 Sol,GPT-5.6,active,USD,1M tokens,5.0,30.0,0.5,6.25,2.5,15.0,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,gpt-5.6-terra,GPT-5.6 Terra,GPT-5.6,active,USD,1M tokens,2.5,15.0,0.25,3.125,1.25,7.5,https://developers.openai.com/api/docs/pricing,2026-07-10T00:00:00Z,2026-07-10T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
-openai,o3,o3,o-series,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
-xai,grok-4.3,Grok 4.3,Grok 4,active,USD,1M tokens,1.25,2.5,,,,,https://docs.x.ai/developers/models,2026-07-05T00:00:00Z,2026-07-05T00:00:00Z,,xAI official model page lists Grok 4.3 input and output rates per 1M tokens. Cached and batch rates were not shown in the extracted official model card.
+anthropic,claude-haiku-4.5,Claude Haiku 4.5,Claude Haiku,active,USD,1M tokens,1.0,5.0,0.1,1.25,0.5,2.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,cache_write stores the 5-minute write rate.
+anthropic,claude-opus-4.8,Claude Opus 4.8,Claude Opus,active,USD,1M tokens,5.0,25.0,0.5,6.25,2.5,12.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"Anthropic table columns are input, 5-minute cache write, 1-hour cache write, cache read, and output. cache_write stores the 5-minute write rate."
+anthropic,claude-sonnet-4.6,Claude Sonnet 4.6,Claude Sonnet,active,USD,1M tokens,3.0,15.0,0.3,3.75,1.5,7.5,https://docs.anthropic.com/en/docs/about-claude/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,cache_write stores the 5-minute write rate.
+anthropic,claude-sonnet-5-intro,Claude Sonnet 5 introductory pricing,Claude Sonnet,active,USD,1M tokens,2.0,10.0,0.2,2.5,1.0,5.0,https://www.anthropic.com/news/claude-sonnet-5,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,2026-06-30,Introductory pricing applies through 2026-08-31 and standard pricing starts 2026-09-01. V1 must not switch early to the standard $3/M input and $15/M output rates.
+cohere,aya-expanse-32b,Aya Expanse 32B,Aya,active,USD,1M tokens,0.5,1.5,,,,,https://cohere.com/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Cohere official pricing page states Aya Expanse 8B and 32B API pricing is $0.50/1M input and $1.50/1M output.
+cohere,command-a-plus,Command A+,Command,active,USD,1M tokens,,,,,,,https://docs.cohere.com/docs/command-a-plus,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"Cohere's official Command A+ page states that access is free until applicable rate limits are reached and that production use is available through Model Vault. Another official Cohere page shows per-token pricing for the same model ID, so V1 keeps token pricing null until the official documentation is reconciled."
+cohere,command-r-plus-08-2024,Command R+ 08-2024,Command R,active,USD,1M tokens,2.5,10.0,,,,,https://cohere.com/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,2024-08-01,Cohere official pricing page lists Command R+ 08-2024 at $2.50/1M input tokens and $10.00/1M output tokens.
+deepseek,deepseek-v4-flash,DeepSeek V4 Flash,DeepSeek V4,active,USD,1M tokens,0.14,0.28,0.0028,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
+deepseek,deepseek-v4-pro,DeepSeek V4 Pro,DeepSeek V4,active,USD,1M tokens,0.435,0.87,0.003625,,,,https://api-docs.deepseek.com/quick_start/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,DeepSeek labels input cache miss as input price and cache hit as cached_input.
+google-gemini,gemini-2.5-flash,Gemini 2.5 Flash,Gemini 2.5,active,USD,1M tokens,0.3,2.5,0.03,,0.15,1.25,https://ai.google.dev/gemini-api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Rates are for text/image/video input; Google lists separate audio rates in the official pricing table.
+google-gemini,gemini-2.5-pro,Gemini 2.5 Pro,Gemini 2.5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://ai.google.dev/gemini-api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,Rates are for prompts up to 200k tokens. Google lists higher rates above 200k tokens in the official pricing table.
+mistral-ai,mistral-large,Mistral Large 3,Mistral Large,active,USD,1M tokens,0.5,1.5,0.05,,0.25,0.75,https://mistral.ai/pricing/api/,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,mistral-large-latest currently maps to Mistral Large 3. Standard price is $0.50/M input and $1.50/M output. Cached input receives a 90% discount. Batch processing receives a 50% discount.
+openai,gpt-4.1,GPT-4.1,GPT-4.1,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5,GPT-5,GPT-5,active,USD,1M tokens,1.25,10.0,0.125,,0.625,5.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5.4-mini,GPT-5.4 mini,GPT-5.4,active,USD,1M tokens,0.75,4.5,0.075,,0.375,2.25,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+openai,gpt-5.5,GPT-5.5,GPT-5.5,active,USD,1M tokens,5.0,30.0,0.5,,2.5,15.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens. Batch prices are 50% of the listed standard price where shown.
+openai,gpt-5.6-luna,GPT-5.6 Luna,GPT-5.6,active,USD,1M tokens,1.0,6.0,0.1,1.25,0.5,3.0,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,gpt-5.6-sol,GPT-5.6 Sol,GPT-5.6,active,USD,1M tokens,5.0,30.0,0.5,6.25,2.5,15.0,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,gpt-5.6-terra,GPT-5.6 Terra,GPT-5.6,active,USD,1M tokens,2.5,15.0,0.25,3.125,1.25,7.5,https://developers.openai.com/api/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,"OpenAI official pricing page lists Standard, Batch, Flex, and Priority rates per 1M tokens. Public V1 fields store Standard short-context prices and Batch short-context prices."
+openai,o3,o3,o-series,active,USD,1M tokens,2.0,8.0,0.5,,1.0,4.0,https://platform.openai.com/docs/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,OpenAI official pricing page lists standard and batch rates per 1M tokens.
+xai,grok-4.3,Grok 4.3,Grok 4,active,USD,1M tokens,1.25,2.5,0.2,,1.0,2.0,https://docs.x.ai/developers/pricing,2026-07-27T00:00:00Z,2026-07-27T00:00:00Z,,V1 stores short-context pricing. The long-context threshold is 200K tokens. Grok 4.3 batch pricing receives a 20% discount. Long-context prices are not stored in the current short-context fields.