Mach-1-Additive-35B / README.md
maniac-11111's picture
final
9d3ee31
|
Raw
History Blame Contribute Delete
1.96 kB
<!-- Theme-aware embed. If your renderer ignores <picture>, use the plain
image line underneath instead and delete this block. -->
![Screenshot 2026-08-03 at 10.32.36 AM](https://cdn-uploads.huggingface.co/production/uploads/680868b984cac4b136ea7900/0YxMKj31WMJT1NwbVAGjj.png)
<!-- Plain fallback:
![Capability retention vs BF16 teacher](assets/benchmarks-light.png)
-->
## Benchmarks
| | Mean retention, 12 benchmarks |
| --- | ---: |
| **Mach-1 Small** | **95.0%** |
| Ternary Bonsai 27B (PrismML) | 93.6% |
| Gemma 4 Q2_K_XL (Unsloth) | 85.6% |
### Per-benchmark Retention
| Benchmark | Mach-1 Small | Ternary Bonsai 27B | Gemma 4 Q2_K_XL |
| --- | ---: | ---: | ---: |
| AIME26 | **99.5%** | 92.7% | 67.7% |
| MATH-500 | **99.4%** | 98.2% | 95.6% |
| AIME25 | **99.1%** | 91.7% | 67.2% |
| GSM8K | 98.4% | **100.2%** | 97.3% |
| MBPP+ | 98.3% | **98.4%** | 92.2% |
| HumanEval+ | 97.4% | **98.7%** | 94.1% |
| MMLU-Redux | 96.2% | 94.0% | **96.9%** |
| IFEval | 94.0% | 89.8% | **95.5%** |
| MuSR | **92.7%** | 91.6% | 91.1% |
| BFCL-v3 | 92.0% | **98.9%** | 95.7% |
| τ²-bench | 90.0% | **91.2%** | 73.1% |
| IFBench | **83.2%** | 77.7% | 61.3% |
| **Mean** | **95.0%** | **93.6%** | **85.6%** |
Mach-1 Small's own scores and teacher scores:
| Benchmark | Score | Teacher (Qwen3.6-35B-A3B BF16) | Retention |
| --- | ---: | ---: | ---: |
| AIME26 | 89.58 | 90.00 | 99.5% |
| MATH-500 | 98.00 | 98.60 | 99.4% |
| AIME25 | 87.50 | 88.33 | 99.1% |
| GSM8K | 94.69 | 96.21 | 98.4% |
| MBPP+ | 94.44 | 96.03 | 98.3% |
| HumanEval+ | 92.68 | 95.12 | 97.4% |
| MMLU-Redux | 89.18 | 92.68 | 96.2% |
| IFEval | 83.75 | 89.05 | 94.0% |
| MuSR | 61.77 | 66.66 | 92.7% |
| BFCL-v3 | 68.97 | 74.98 | 92.0% |
| τ²-bench | 71.58 | 79.51 | 90.0% |
| IFBench | 54.08 | 64.97 | 83.2% |
### Speed
![Screenshot 2026-08-03 at 11.55.47 AM](https://cdn-uploads.huggingface.co/production/uploads/680868b984cac4b136ea7900/lYFfUCU1QWrGjM23vbb66.png)