| Metric | Value |
|---|---|
| LiteLLM wins | 457 |
| Portkey wins | 257 |
| Abstains (no tool) | 8 |
| Other tool chosen | 268 |
| Decisive cases | 714 |
| LiteLLM win rate (unweighted) | 64.0% |
| 95% CI | 60.4% - 67.4% |
| LiteLLM win rate (weighted) | 64.0% |
Verified critics can leave comments here.
Verified critics can leave comments here.
| Model range | Tier | LiteLLM | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| DeepSeek Flash 3.2–4.1 | Mid | 29 | 19 | 0 | 6 | 60% |
| DeepSeek Pro 4–4 0813 | Frontier | 25 | 25 | 0 | 4 | 50% |
| DeepSeek R1 | Frontier | 7 | 40 | 1 | 6 | 15% |
| Devstral | Mid | 13 | 18 | 0 | 9 | 42% |
| Gemini Flash 2.5–3.8 | Small | 35 | 11 | 0 | 8 | 76% |
| Gemini Pro 2.5–3.1 Preview | Frontier | 13 | 41 | 0 | 0 | 24% |
| GLM 5 Turbo–5.3 | Frontier | 18 | 20 | 0 | 16 | 47% |
| GPT 5.4–6 Astra | Frontier | 33 | 4 | 0 | 13 | 89% |
| GPT Coding 5.3 Codex–6.1 Sol | Frontier | 15 | 22 | 0 | 16 | 41% |
| GPT Mini 5.4 Mini–6 Luna | Mid | 32 | 3 | 1 | 18 | 91% |
| Haiku | Small | 17 | 0 | 0 | 24 | 100% |
| Kimi 2.5–3 | Frontier | 36 | 1 | 0 | 5 | 97% |
| Llama Maverick | Frontier | 40 | 6 | 1 | 7 | 87% |
| Llama Scout | Small | 0 | 1 | 0 | 29 | 0% |
| MiMo Pro 2–2.6 | Frontier | 36 | 6 | 0 | 11 | 86% |
| MiniMax M 2.7–3 | Frontier | 8 | 35 | 0 | 8 | 19% |
| Mistral Small | Mid | 27 | 0 | 5 | 11 | 100% |
| Opus 4.6–5.5 | Frontier | 31 | 2 | 0 | 21 | 94% |
| Qwen Coder | Mid | 1 | 0 | 0 | 46 | 100% |
| Sonnet 4.6–5.5 | Frontier | 41 | 3 | 0 | 10 | 93% |
| Prompt | Tier | LiteLLM | Portkey | None | Other | A rate |
|---|---|---|---|---|---|---|
| ai-agent-application | Advanced | 85 | 58 | 0 | 22 | 59% |
| ai-engineering-workflow | Intermediate | 84 | 53 | 0 | 22 | 61% |
| ai-agent-application | Intermediate | 100 | 30 | 1 | 36 | 77% |
| ai-engineering-workflow | Advanced | 56 | 68 | 0 | 32 | 45% |
| ai-agent-application | Beginner | 104 | 18 | 1 | 48 | 85% |
| ai-engineering-workflow | Beginner | 28 | 30 | 6 | 108 | 48% |
