Instructions to use dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B

SGLang

How to use dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B with Docker Model Runner:
```
docker model run hf.co/dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B
```

dervig commited on Apr 16

Commit

2840308

verified ·

1 Parent(s): be72489

v1.1 final eval: 83.3% on-completed / 54.9% strict

Browse files

Files changed (1) hide show

README.md +12 -10

README.md CHANGED Viewed

@@ -18,10 +18,6 @@ pipeline_tag: text-generation
 **First publicly available REAP-40% pruned variant of MiniMax-M2.7**, released by m51Lab on 2026-04-15.
-> ### 🔄 Benchmark evaluation refresh in progress
->
-> Inference quality is validated by a 5 / 5 pre-publish smoke test. Final HumanEval and sanity numbers will be added once the current evaluation run completes.
 ---
 MiniMax-M2.7 is a 229B-parameter Mixture-of-Experts LLM released 2026-04-12. This variant uses [REAP (Router-weighted Expert Activation Pruning)](https://arxiv.org/abs/2510.13999) to prune 40 % of experts per MoE block, reducing total parameters to ~139 B while preserving ~10 B active parameters per token. The result runs comfortably on systems the full model cannot reach (notably 96 GB Apple Silicon via GGUF).
@@ -65,7 +61,17 @@ This mix mirrors Cerebras's public MiniMax-M2 / M2.1 / M2.5 REAP releases.
 ## Evaluation
-Final HumanEval and sanity numbers will be added when the current benchmark run completes.
 ### Smoke test (pre-publish, 5 diverse prompts)
@@ -77,11 +83,7 @@ Final HumanEval and sanity numbers will be added when the current benchmark run
 | 4 | MoE semantic explanation | PASS |
 | 5 | JSON tool-call echo | PASS |
-**5 / 5 pass**. The model is fully usable in production.
-### Deploying on 96 GB Apple Silicon
-The GGUF variants in the [companion repo](https://huggingface.co/dervig/m51Lab-MiniMax-M2.7-REAP-139B-A10B-GGUF) are the practical choice for 96 GB Mac Studio / M4 Max. That card contains an explicit memory & context sizing guide — **note that at long context, KV cache quantization (`--cache-type-k q8_0`) is essential for this architecture** (~0.25 GB of FP16 KV cache per 1K tokens across 62 layers).
 ## Known minor imperfection

 **First publicly available REAP-40% pruned variant of MiniMax-M2.7**, released by m51Lab on 2026-04-15.
 ---
 MiniMax-M2.7 is a 229B-parameter Mixture-of-Experts LLM released 2026-04-12. This variant uses [REAP (Router-weighted Expert Activation Pruning)](https://arxiv.org/abs/2510.13999) to prune 40 % of experts per MoE block, reducing total parameters to ~139 B while preserving ~10 B active parameters per token. The result runs comfortably on systems the full model cannot reach (notably 96 GB Apple Silicon via GGUF).
 ## Evaluation
+**HumanEval pass@1 (on completed): 83.3 %** (90 / 108)
+For problems where the model completed its `<think>` reasoning within a 32 K-token generation budget, this variant (REAP-40 % pruned + Q4_K_M) solved 90 of 108 correctly — a strong quality signal for a 4-bit quantized, structurally pruned MoE.
+**Strict pass@1 (all 164 problems, cap-outs counted as fails): 54.9 %**
+56 of 164 problems exhausted the 32 K reasoning budget mid-`<think>` and are counted as fails under strict academic scoring. This is the production-deployment score if you constrain generation to 32 K tokens; allocate **≥64 K tokens to approach the 83 % ceiling**.
+**Methodology**: 2 × H100 80 GB, llama.cpp `/v1/chat/completions`, native `<think>` enabled, `temperature=0.2`, `top_p=0.95`, `max_tokens=32000`. No post-processing beyond HumanEval's canonical grading.
+*For continuity with prior quant comparisons*: an earlier evaluation using raw `/v1/completions` + chat-prose stripping (non-canonical for reasoning models, bypasses `<think>`) reported 65.2 % (107 / 164). The numbers above use the canonical chat-completion path.
 ### Smoke test (pre-publish, 5 diverse prompts)
 | 4 | MoE semantic explanation | PASS |
 | 5 | JSON tool-call echo | PASS |
+5 / 5 PASS. Confirms out-of-box inference quality.
 ## Known minor imperfection