Hermes 4 – Llama-3.1 70B (Reasoning)
Analysis Summary
Hermes 4 on Llama-3.1 70B (Reasoning) is the extended-thinking variant of Nous Research's 70B fine-tune, and the reasoning mode delivers a substantial uplift on math and coding benchmarks relative to its non-reasoning sibling. A math index of 68.7, GPQA of 0.699, and LiveCodeBench of 0.653 are strong for a 70B-class model, and MMLU-Pro of 0.811 is competitive.
For businesses, this model offers a cost-efficient path to reasoning-capable performance. At $0.13/$0.40 per million tokens, it is among the more affordable options for tasks requiring deliberate multi-step thinking. Its agentic index is low, limiting its use in tool-calling or autonomous workflows, and long-context reliability is weak.
Teams with budget constraints who need reasoning capability for analytical, math-heavy, or coding tasks will find this a strong value proposition, though the 405B reasoning variant offers additional headroom for the most demanding workloads.
Assessed June 6, 2026
Editorial notes
Hermes 4 Llama-3.1 70B (Reasoning) from Nous Research delivers strong math performance and a LiveCodeBench of 0.653 at a very low price of $0.13/$0.40 per million tokens.
Rankings consider pricing, capabilities, benchmarks, and real-world applicability and are refreshed as new models launch. Feedback?
Performance Profile
How Hermes 4 – Llama-3.1 70B (Reasoning) compares
Hermes 4 – Llama-3.1 70B (Reasoning) ranks #215 of 378 AI models we track for overall intelligence, #188 of 315 for coding, #233 of 289 for agentic tasks. At $0.13 per million input tokens it is cheaper than 59% of comparable models.
Performance Indices
Source: Artificial Analysis
Benchmark Scores
Intelligence
Technical
Content
Benchmark data from Artificial Analysis and Hugging Face
How does Hermes 4 – Llama-3.1 70B (Reasoning) stack up?
Compare side-by-side with other professional models.
Model Information
| Provider | Nous Research |
| Release Date | August 27, 2025 |
| Status | Active |
Pricing
| Token Type | Cost per 1M tokens | Cost per 1K tokens |
|---|---|---|
| Input | $0.13 | $0.000130 |
| Output | $0.40 | $0.000400 |
Leaderboard Categories
Frequently asked questions about Hermes 4 – Llama-3.1 70B (Reasoning)
How much does Hermes 4 – Llama-3.1 70B (Reasoning) cost?
Hermes 4 – Llama-3.1 70B (Reasoning) costs $0.13 per million input tokens and $0.40 per million output tokens.
Is Hermes 4 – Llama-3.1 70B (Reasoning) good for coding?
On our coding benchmark index, Hermes 4 – Llama-3.1 70B (Reasoning) ranks #188 of 315 models, placing it in the broader range of the field for code generation and debugging.
Who created Hermes 4 – Llama-3.1 70B (Reasoning)?
Hermes 4 – Llama-3.1 70B (Reasoning) is developed by Nous Research and was released on August 27, 2025.
