Nous Research · Hermes 4
Hermes 4 70B
Nous Research's open-weights hybrid reasoning fine-tune of Llama 3.1 70B.
- AA Intelligence
- 7.9#112 of 142
- Price
- Self-host
- Output speed
- —
- Context window
- 128K tokens
Composite indexes
- Artificial Analysis Intelligence Index#112 of 1428
Leader: Claude Opus 5.5 at 58
Coding
- LiveCodeBench#30 of 5665.3%
Leader: Fugu at 92.9%
Agents and tool use
- τ²-Bench#63 of 8822.5%
Leader: GLM-5.2 at 99.1%
Reasoning and knowledge
- GPQA Diamond#98 of 14269.9%
Leader: GPT-6 Astra at 96.1%
- Humanity’s Last Exam#109 of 1478.8%
Leader: Claude Opus 5.5 at 61.4%
- CritPt#138 of 1390%
Leader: GPT-5.6 Sol at 32.3%
Vision and long context
- AA Long Context Reasoning#125 of 1399.7%
Leader: Kimi K3 at 88.7%
Other published results
| Result | Score | Reported by | Source |
|---|---|---|---|
| aime-2025 · AIME 2025; reasoning | 68.7% | independent | Artificial Analysis: Hermes 4 - Llama-3.1 70B (Reasoning) ↗ |
| Terminal-Bench Hard (AA) · reasoning | 4.5% | independent | Artificial Analysis: Hermes 4 - Llama-3.1 70B (Reasoning) ↗ |
| IFBench · reasoning | 31.3% | independent | Artificial Analysis: Hermes 4 - Llama-3.1 70B (Reasoning) ↗ |
| AA-Omniscience Index · reasoning | -49.4 index | independent | Artificial Analysis: Hermes 4 - Llama-3.1 70B (Reasoning) ↗ |