The Inference Report
The SWE-rebench rankings remain frozen at their previous positions, with AnthropicFable 5 holding 64.5% ± 1.41% at the top and the full top seventeen models showing identical scores and confidence intervals to the prior snapshot. On the Artificial Analysis benchmark, a single structural change has occurred: K-EXAONE 2.0 enters at position 48 with a score of 7.6, pushing Claude 4.5 Haiku from 48 to 49 and cascading all subsequent models down one position through the bottom of the ranking. This insertion does not reflect any score movement among existing models but rather adds a new entrant to the tracked set. The stability across SWE-rebench suggests either that the evaluation window has not cycled or that the variance bands (ranging from ± 0.54% to ± 1.83% in the top tier) contain any real performance shifts. On Artificial Analysis, the new entry at the lower end of the ranking is the only movement; the top performers including Gemini 4 Carbon at 60 and Claude Opus 5.5 at 59.1 remain unchanged. Neither benchmark shows evidence of meaningful reordering among models already in the rankings, making today's data primarily a confirmation of prior standings with one new data point in a secondary benchmark.
Cole Brennan
The Value FrontierDaily rankings from SWE-rebench, a benchmark designed to fairly compare LLM capabilities on real-world software engineering tasks. Unlike other evaluations, it uses a standardized scaffolding for all models, continuously updates its dataset to prevent contamination, and runs each model five times to account for stochastic variance.
| # | Model | Score |
|---|---|---|
| 1 | AnthropicFable 5 [high]Model | 64.5%± 1.41% |
| 2 | GrokGrok 4.5 [high]Model | 63.8%± 0.60% |
| 3 | AnthropicOpus 5 [high]Model | 63.4%± 1.35% |
| 4 | Z.aiGLM-5.2 [high]Model | 62.9%± 1.19% |
| 5 | OpenAIGPT-5.6 Sol [medium]Model | 62.3%± 1.83% |
| 6 | JunieJunieAgent | 61.8%± 0.54% |
| 7 | AnthropicClaude CodeAgent | 60.4%± 1.03% |
| 8 | OpenAICodexAgent | 58.0%± 1.29% |
| 9 | AnthropicSonnet 5 [high]Model | 56.8%± 0.94% |
| 10 | CursorCursorAgent | 51.7%± 0.84% |
Artificial Analysis composite index across coding, math, and reasoning benchmarks.
| # | Model | Score | tok/s | $/1M |
|---|---|---|---|---|
| 1 | Gemini 4 Carbon | 60 | 91 | $8.00 |
| 2 | Claude Opus 5.5 | 59.1 | 95 | $8.00 |
| 3 | GPT-6 Astra | 56.9 | 46 | $20.00 |
| 4 | GPT-6.1 Sol | 54.6 | 60 | $4.00 |
| 5 | Claude Sonnet 5.5 | 54.5 | 136 | $4.00 |
| 6 | GPT-6.1 Terra | 51.3 | 114 | $2.00 |
| 7 | Gemini 4 Argon | 50.4 | 89 | $8.00 |
| 8 | Muse Wonder Full | 50.1 | — | — |
| 9 | Claude Fable 5.1 | 50 | 61 | $20.00 |
| 10 | Kimi K3.1 Pro | 46.5 | 27 | $4.00 |
Output tokens per second — higher is faster. Minimum intelligence score of 40.
| # | Model | tok/s |
|---|---|---|
| 1 | Claude Sonnet 5.5 | 136 |
| 2 | Muse Wonder | 134 |
| 3 | GPT-6 Sol | 127 |
| 4 | GPT-6.1 Terra | 114 |
| 5 | Claude Opus 5.5 | 95 |
| 6 | Gemini 4 Carbon | 91 |
| 7 | Gemini 4 Argon | 89 |
| 8 | Claude Fable 5.1 | 61 |
| 9 | GPT-6.1 Sol | 60 |
| 10 | Claude Opus 5 | 50 |
Blended cost per 1M tokens (3:1 input/output) — lower is cheaper. Minimum intelligence score of 40.
| # | Model | $/1M |
|---|---|---|
| 1 | GPT-6.1 Terra | $2.00 |
| 2 | GPT-6.1 Sol | $4.00 |
| 3 | Claude Sonnet 5.5 | $4.00 |
| 4 | Kimi K3.1 Pro | $4.00 |
| 5 | GPT-6 Sol | $4.00 |
| 6 | Gemini 4 Carbon | $8.00 |
| 7 | Claude Opus 5.5 | $8.00 |
| 8 | Gemini 4 Argon | $8.00 |
| 9 | Muse Wonder | $8.00 |
| 10 | Claude Opus 5 | $10.00 |