Leading: Portkey (63.9%)
| Metric | Value |
|---|---|
| OpenRouter wins | 115 |
| Portkey wins | 204 |
| Abstains (no tool) | 6 |
| Other tool chosen | 326 |
| Decisive cases | 319 |
| OpenRouter win rate (unweighted) | 36.1% |
| 95% CI | 31.0% - 41.5% |
| OpenRouter win rate (weighted) | 36.1% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model | Tier | OpenRouter | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| Gemini 2.5 Pro | Frontier | 0 | 35 | 0 | 1 | 0% |
| MiniMax M3 | Frontier | 6 | 25 | 0 | 4 | 19% |
| GLM 5.2 | Frontier | 13 | 17 | 0 | 6 | 43% |
| GPT 5.3 Codex | Frontier | 11 | 19 | 0 | 5 | 37% |
| DeepSeek R1 0528 | Frontier | 2 | 26 | 1 | 7 | 7% |
| DeepSeek V4 Pro | Frontier | 2 | 21 | 0 | 13 | 9% |
| Claude Opus 4.8 | Frontier | 16 | 2 | 0 | 18 | 89% |
| DeepSeek V4 Flash | Mid | 2 | 16 | 0 | 18 | 11% |
| Qwen3 Coder Next | Mid | 16 | 0 | 0 | 15 | 100% |
| Gemini 3.5 Flash | Small | 5 | 10 | 0 | 21 | 33% |
| Devstral 2 2512 | Mid | 0 | 15 | 0 | 13 | 0% |
| GPT 5.4 Mini | Mid | 13 | 0 | 1 | 22 | 100% |
| MiMo V2.5 Pro | Frontier | 8 | 5 | 0 | 23 | 62% |
| GPT 5.5 | Frontier | 8 | 4 | 0 | 21 | 67% |
| Claude Sonnet 4.6 | Frontier | 6 | 2 | 0 | 28 | 75% |
| Llama 4 Maverick | Frontier | 0 | 5 | 1 | 30 | 0% |
| Claude Haiku 4.5 | Small | 4 | 0 | 0 | 24 | 100% |
| Kimi K2.7 Code | Frontier | 3 | 1 | 0 | 21 | 75% |
| Llama 4 Scout | Small | 0 | 1 | 0 | 13 | 0% |
| Mistral Small 4 | Mid | 0 | 0 | 3 | 23 | n/a |
| Prompt | Tier | OpenRouter | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-engineering-workflow | Beginner | 64 | 24 | 4 | 20 | 73% |
| ai-engineering-workflow | Advanced | 1 | 56 | 0 | 47 | 2% |
| ai-agent-application | Intermediate | 22 | 25 | 1 | 63 | 47% |
| ai-agent-application | Advanced | 3 | 43 | 0 | 60 | 7% |
| ai-engineering-workflow | Intermediate | 2 | 44 | 0 | 58 | 4% |
| ai-agent-application | Beginner | 23 | 12 | 1 | 78 | 66% |
