Nous ResearchHermes 4 - Llama-3.1 405B
Hermes 4 - Llama-3.1 405B is a Nous Research open-weights model, released 27 Aug 2025. Its best configuration (no reasoning reasoning effort) ranks #567 of 785 on the BenchLeader Index at 43.4 ±3.0, in the lower half. The ± is the point: 288 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in knowledge (47) and lowest in long context (35). At $1.50 per million tokens blended it is pricier than most ranked models. Output speed of 40 tokens per second puts it in the slowest quarter, with a first answer in 2.3 s. It has been measured at 2 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 27 Aug 2025.stale: no new result in six months
- Blended price
- $1.50/M
- $1.00 in · $3.00 out
- Output speed
- 40 tok/s
- measured by Artificial Analysis
- First answer
- 2.28 s
- first token 2.28 s
- Context
- 128k
- Full answer
- 15 s
- median, reasoning included
- Cost per run
- –
- Released
- 27 Aug 2025
- Overall index43
- Reasoning41
- Agents & tools42
- Knowledge47
- Instruction following38
- Long context35
- Composite38
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoningbest | 43.4 | #567 | 40 tok/s | 2.28 s | $0.0013 | Agents & tools 42 · Composite 38 · Instruction following 38 · Knowledge 47 · Long context 35 · Reasoning 41 |
| thinking | 43.3 | #570 | 41 tok/s | 51 s | $0.0013 | Agents & tools 44 · Composite 38 · Instruction following 36 · Knowledge 45 · Long context 35 · Reasoning 41 |
How its index has moved
9 Sept 2026 to 21 Oct 2026- no reasoning
- thinking
The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| GPQA Diamond (AA)not in index | 53.6%#410 | 72.7%#274 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 4.2%#522 | 10.9%#307 | Artificial Analysis |
| CritPt | 0.0%#315 | 0.3%#271 | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| Terminal-Bench Hard | 9.8%#228 | 11.4%#215 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 26.6%#280 | 22.2%#314 | Artificial Analysis |
Knowledge
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| AA-Omniscience | -33.3#295 | -36.0#304 | Artificial Analysis |
| AA-Omniscience: accuracynot in index | 26.1%#220 | 30.1%#182 | Artificial Analysis |
| AA-Omniscience: non-hallucinationnot in index | 19.8%#282 | 5.5%#517 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| IFBench | 34.8%#313 | 32.7%#338 | Artificial Analysis |
Long context
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| AA-LCR | 22.0%#424 | 22.3%#422 | Artificial Analysis |
Composite
| Benchmark | no reasoning | thinking | Source |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 7.4#477 | 7.5#472 | Artificial Analysis |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0013 | – | 9.7 s |
| Summarize a 30-page report | 12,000 / 600 | $0.014 | – | 17.2 s |
| Code edit | 6,000 / 1,500 | $0.011 | – | 39.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.072 | – | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.0026 | – | 7.3 s |
More from Nous Research
See also
Questions people ask
- What are Hermes 4 - Llama-3.1 405B's benchmark scores?
- As of 25 Oct 2026, Hermes 4 - Llama-3.1 405B (no reasoning effort) scores 43.4 ±3.0 on the BenchLeader Index, strongest in knowledge (47) and weakest in long context (35). That rests on 6 independent benchmark results, and every score on this page links to its publisher.
- How does Hermes 4 - Llama-3.1 405B rank?
- Hermes 4 - Llama-3.1 405B (no reasoning effort) is #567 of 785 ranked configurations on 25 Oct 2026. 288 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Hermes 4 - Llama-3.1 405B cost per 1M tokens?
- As of 25 Oct 2026, Hermes 4 - Llama-3.1 405B costs $1.00 per 1M input tokens and $3.00 per 1M output tokens, $1.50 blended (three input tokens per output token).
- When was Hermes 4 - Llama-3.1 405B released?
- Nous Research released Hermes 4 - Llama-3.1 405B on 27 Aug 2025. BenchLeader has results for it at 2 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Hermes 4 - Llama-3.1 405B: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/hermes-4-llama-3-1-405b.