Compare / Hermes 4.3 36B vs Kimi Linear 48B-A3B
Hermes 4.3 36B vs Kimi Linear 48B-A3B
Updated November 2026
Verdict
Hermes 4.3 36B scores higher on the Small LM Index: 53.2 vs 36.6. On the 3 benchmarks both labs report, Hermes 4.3 36B wins 3 and Kimi Linear 48B-A3B wins 0. Hermes 4.3 36B is smaller: it needs about 25.3 GB of memory at 4-bit, against 30.8 GB for Kimi Linear 48B-A3B. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Hermes 4.3 36B if…
- you want the strongest model overall
- you need math and reasoning
- you have limited memory (about 25.3 GB at 4-bit)
Pick Kimi Linear 48B-A3B if…
- you work with long documents
- you want the more widely used model (more quantized builds and community support)
Specifications
| Hermes 4.3 36B | Kimi Linear 48B-A3B | |
|---|---|---|
| Small LM Index | 53.2 (likely 50–56) | 36.6 (likely 34–40) |
| Lab | Nous Research (Hermes) · United States | Moonshot AI (Kimi) · China |
| Released | 2025-11-17 | 2025-10-30 |
| Parameters | 36.2B (dense) | 48B (3B active, MoE) |
| Memory at 4-bit | ~25.3 GB | ~30.8 GB |
| Context window | 512K tokens | 1M tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | No thinking |
| License | Apache-2.0 | MIT |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 2.2K | 69K |
Benchmarks
Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.
| Benchmark | Hermes 4.3 36B | Kimi Linear 48B-A3B |
|---|---|---|
| MMLU-Pro Knowledge | 80.7 | 58.5✓ |
| GPQA Diamond Knowledge | 65.5 | 41.2✓ |
| Humanity's Last Exam Reasoning | ≈ 9 | 2.5✓ |
| AIME Math | 69.3 | 36.3✓ |
| LiveCodeBench Coding | ≈ 56 | 37.8✓ |
| τ²-Bench Agentic | — | 0.0✓ |
| IFBench Instruction following | — | 28.1✓ |
| MMLU Knowledge | 87.7 | — |
| IFEval Instruction following | 77.9 | — |
| MATH-500 Math | 93.8 | — |
| AA-LCR Long context | — | 28.0✓ |
Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.
Questions
Which is better, Hermes 4.3 36B or Kimi Linear 48B-A3B?
Hermes 4.3 36B scores higher on the Small LM Index: 53.2 vs 36.6. On the 3 benchmarks both labs report, Hermes 4.3 36B wins 3 and Kimi Linear 48B-A3B wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Hermes 4.3 36B is smaller: it needs about 25.3 GB of memory at 4-bit, against 30.8 GB for Kimi Linear 48B-A3B.
Which has the longer context window?
Kimi Linear 48B-A3B has the longer context window: 1M tokens against 512K.
Can I use Hermes 4.3 36B and Kimi Linear 48B-A3B commercially?
Hermes 4.3 36B is released under Apache-2.0 and Kimi Linear 48B-A3B under MIT. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.