Update README.md

README.md

@@ -37,12 +37,14 @@ Two variants are provided, differing only in *how many* of the later transformer
 
 | File | Blocks taken from `ref2va` | Tradeoff |
 |---|---|---|
-| `minimax_h3_hybrid_fl2va_ref2va_b30-49.safetensors` | 30–49 (last 20 of 50) | Closer to `fl2va` — higher visual/audio quality, slightly reduced reference capability. |
+| `minimax_h3_hybrid_fl2va_ref2va_b30-49.safetensors` | 30–49 (last 20 of 50) | Closer to `fl2va` — higher visual/audio quality, somewhat reduced reference capability. |
+| `minimax_h3_hybrid_fl2va_ref2va_b25-49.safetensors` | 25–49 (last 25 of 50) | Closer to `fl2va` — slight higher visual/audio quality, slightly reduced reference capability. |
 | `minimax_h3_hybrid_fl2va_ref2va_b20-49.safetensors` | 20–49 (last 30 of 50) | Closer to `ref2va` — slightly higher reference capability, slightly lower visual/audio quality. |
+| `minimax_h3_hybrid_fl2va_ref2va_b15-49.safetensors` | 15–49 (last 35 of 50) | Closer to `ref2va` — higher reference capability, somewhat lower visual/audio quality. |
 
 Both are reasonable defaults depending on what you're optimizing for: pick **b30-49** if output fidelity matters more and reference conditioning just needs to work well; pick **b20-49** if faithfulness to the reference input is the priority and you can tolerate a small step down in raw quality. Neither variant is uniformly better than the other — the difference is a fairly narrow, gradual tradeoff rather than a clear win/loss.
 
-I personally recommend trying the *b30-49* version first, and if you find you need stronger reference adherence (at the cost of output quality), try the *b20-49* version.
+I personally recommend trying the *b25-49* version first, and if you find you need stronger reference adherence (at the cost of output quality), try the *b20-49* or *b15-49* version.
 
 ## What this model is