Model Comparison
Hermes 3 - Llama-3.1 70B
vs. Hermes 4 - Llama-3.1 405B (Non-reasoning)
Comparing 2 AI models · 12 benchmarks · Nous Research
Recommended Pick
No Hermes 4 - Llama-3.1 405B (Non-reasoning) 7 metric wins Strongest on: Throughput, Intelligence, GPQA
Best Value
Hermes 3 - Llama-3.1 70B
100.0 value score
22.2 reasoning / $0.30/1M
Lowest Price
Hermes 3 - Llama-3.1 70B
$0.30/1M input price
Best Reasoning
Hermes 3 - Llama-3.1 70B
22.2 reasoning score
Blends available reasoning benchmarks
Best for Coding
Hermes 4 - Llama-3.1 405B (Non-reasoning)
18.1 coding index
Composite Indices
Higher is better; speed and price are normalized
Standard Benchmarks
Only benchmarks with data are shown
Differences That Matter
Best value
Hermes 3 - Llama-3.1 70B has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
Hermes 3 - Llama-3.1 70B is 3.3x cheaper on input tokens than Hermes 4 - Llama-3.1 405B (Non-reasoning).
Speed gap
Hermes 4 - Llama-3.1 405B (Non-reasoning) generates about 1.4x as many tokens per second as Hermes 3 - Llama-3.1 70B.
Reasoning gap
Hermes 3 - Llama-3.1 70B leads Hermes 4 - Llama-3.1 405B (Non-reasoning) by 1.0 points on reasoning.
Top-pick rationale
Hermes 4 - Llama-3.1 405B (Non-reasoning) wins 7 measurable categories, including Throughput, Intelligence, GPQA, MMLU Pro.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
Hermes 3 - Llama-3.1 70B
Nous Research
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Hermes 4 - Llama-3.1 405B (Non-reasoning)
Nous Research
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | No Hermes 3 - Llama-3.1 70B | Top Pick No Hermes 4 - Llama-3.1 405B (Non-reasoning) |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $0.30/1M | $1.00/1M |
| Output Cost | $0.30/1M | $3.00/1M |
| Blended (3:1) | $0.30/1M | $1.50/1M |
| Specifications | ||
| Organization | Nous Research | Nous Research |
| Release Date | Aug 15, 2024 | Aug 27, 2025 |
| Performance & Speed | ||
| Throughput | 31.3 tok/s | 42.3 tok/s |
| TTFT | 381ms | 768ms |
| Latency | 381ms | 768ms |
| Composite Indices | ||
| Value Score | 100.0 | 19.1 |
| Reasoning Score | 22.2 | 21.2 |
| Intelligence | 10.6 | 17.6 |
| Coding | — | 18.1 |
| Math | — | 15.3 |
| Standard Benchmarks | ||
| GPQA | 40.1% | 53.6% |
| MMLU Pro | 57.1% | 72.9% |
| HLE | 4.1% | 4.2% |
| LiveCodeBench | 18.8% | 54.6% |
| MATH 500 | 53.8% | — |
| AIME 2025 | — | 15.3% |
| AIME (Original) | 2.3% | — |
| SciCode | 23.1% | 34.6% |
| LCR | — | 20.0% |
| IFBench | — | 34.8% |
| TAU-bench v2 | — | 26.6% |
| TerminalBench Hard | — | 9.8% |
Key Takeaways
Hermes 3 - Llama-3.1 70B offers the best value at $0.30/1M, making it ideal for high-volume applications and cost-conscious projects.
Hermes 3 - Llama-3.1 70B has the strongest reasoning profile with a 22.2 reasoning score, combining the available reasoning-heavy benchmarks.
Hermes 4 - Llama-3.1 405B (Non-reasoning) reaches a 18.1 coding index, making it the top choice for software development and code generation tasks.
All models support context windows of ∞+ tokens, suitable for processing lengthy documents and maintaining extended conversations.
When to Choose Each Model
Hermes 3 - Llama-3.1 70B
- Cost-sensitive applications
- High-volume processing
- Complex reasoning tasks
- Research & analysis
Hermes 4 - Llama-3.1 405B (Non-reasoning)
- Code generation
- Software development
