GPT-5.5 vs Kimi K3
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | GPT-5.5 | Kimi K3 |
|---|---|---|
| Coding | 25.8 GPT-5.5 | 39.5 Kimi K3 |
| Agents | 64.3 GPT-5.5 | 61.9 Kimi K3 |
| Reasoning | 68.4 GPT-5.5 | 55.2 Kimi K3 |
| Knowledge & language | 63.7 GPT-5.5 | 52.9 Kimi K3 |
| Human preference | 53.9 GPT-5.5 | 57.5 Kimi K3 |
Benchmarks both report
GPT-5.5 leads on 17 of the 30 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | GPT-5.5 | Kimi K3 |
|---|---|---|
| Chess Puzzles (Epoch AI) | 63% | 37% |
| LMArena Elo | 1482 | 1488 |
| LMArena WebDev (LMArena) | 1512 | 1656 |
| LMArena Agent (LMArena) | 0.0471 | 0.0108 |
| APEX-Agents (Mercor) | 55.1% | 47.8% |
| DeepSWE v1.1 (Datacurve) | 67% | 68.5% |
| FrontierMath Tier 4 v2 (Epoch AI) | 72.5% | 39% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 85.3% | 72.2% |
| Furniture Assembly (Epoch AI) | 44.2% | 34.2% |
| GPQA Diamond | 90.7% | 93.5% |
| LiveBench Agentic Coding (LiveBench) | 54% | 62.2% |
| LiveBench Coding (LiveBench) | 82.1% | 81.4% |
| LiveBench Data Analysis (LiveBench) | 81.6% | 78.7% |
| LiveBench Instruction Following (LiveBench) | 70.7% | 71.4% |
| LiveBench Language (LiveBench) | 87.4% | 85.5% |
| LiveBench Mathematics (LiveBench) | 95.9% | 84.4% |
| LiveBench Reasoning (LiveBench) | 89.7% | 90.7% |
| MirrorCode (Epoch AI) | 10% | 54.4% |
| Mystery Game Puzzles (Epoch AI) | 56% | 26% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 84.4% | 97.2% |
| SAGE (Vals AI) | 51.5% | 52.8% |
| SimpleQA Verified | 63% | 50.6% |
| Vending-Bench 2 (Andon Labs) | 7523.84 | 5165.04 |
| WeirdML (Håvard Tveit Ihle) | 84.9% | 82.6% |
| AIME 2026 | 100% | 96.7% |
| Toolathlon-Verified (HKUST) | 73.5% | 76.5% |
| tau2-bench Banking Knowledge (Sierra) | 44.6% | 37.1% |
| SimpleBench (SimpleBench) | 69% | 60.7% |
| MMMU-Pro (no tools) | 81.2% | 81.6% |
| OSWorld-Verified | 78.7% | 84.8% |
FAQ
Is GPT-5.5 better than Kimi K3?
GPT-5.5 scores 54.1 on the Respan Index (#23), ahead of Kimi K3 at 53.1 (#25). GPT-5.5 leads on 17 of the 30 benchmarks both report.
Which is cheaper, GPT-5.5 or Kimi K3?
GPT-5.5 lists at $5 input and $30 output per 1M tokens; Kimi K3 at $0.192 and $9.5. Kimi K3 is 4.5x cheaper per token at list price.
Which has the larger context window?
GPT-5.5 accepts 1M tokens and Kimi K3 1.0M tokens.