Model comparison
GPT-5.4 mini vs Kimi K2.6
Kimi K2.6 is the stronger model overall, scoring 47.7 to 45.0 on the Noometry Index.
Last verified . 39 shared benchmarks.
Summary
- They share 39 benchmarks with published results for both. GPT-5.4 mini scores higher in 2 categories and Kimi K2.6 in 8 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where Kimi K2.6 leads 57.0 to 45.5.
- The biggest single-benchmark swing is NYT Connections (extended): 61.8% for GPT-5.4 mini and 87.2% for Kimi K2.6.
- Both cost about the same: $0.75 input and $4.50 output per million tokens.
- GPT-5.4 mini accepts more context: 400K tokens versus 262K.
- Kimi K2.6 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 mini | Kimi K2.6 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 45.0 | 47.7 |
| Released | 2026-03-17 | 2026-04-20 |
| Weights | Proprietary | Open |
| Context window | 400K | 262K |
| Max output | 128K | 262K |
| Input $ / M tokens | $0.75 | $0.95 |
| Output $ / M tokens | $4.50 | $4 |
| Results tracked | 46 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.6 leads
GPT-5.4 mini: 45.2 (#72), Kimi K2.6: 50.7 (#43)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena WebDev | 1397 | 1509 |
| SciCode | 49.9% | 53.5% |
| WeirdML | 60.3% | 55.9% |
| LMArena Coding | 1438 | 1488 |
| ALE-Bench | 1,189 | 1,093 |
| SWE-bench Verified | — | 76.7% |
| FrontierCode | 27% | — |
Agentic & Tool Use GPT-5.4 mini leads
GPT-5.4 mini: 29.9 (#81), Kimi K2.6: 21.9 (#137)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| OSWorld 2.0 | — | 4.6% |
| DeepResearch Bench | 36.3% | — |
| ExploitBench | — | 18.4% |
| GBAEval | — | 0.9% |
| GDP.pdf | — | 12% |
| Vending-Bench 2 | — | 6,205 |
Reasoning Kimi K2.6 leads
GPT-5.4 mini: 30.4 (#85), Kimi K2.6: 40.5 (#55)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| NYT Connections (extended) | 61.8% | 87.2% |
| CritPt | 10% | 8% |
| Chess Puzzles | 24% | 26% |
| LMArena Hard Prompts | 1424 | 1470 |
| Mystery Game Puzzles | 11% | 18% |
| DTBench | 80% | 90.9% |
| LMCA | 40.8% | 37.3% |
| Epoch Capabilities Index | 148.84 | 151.05 |
| ARC-AGI-2 | 18.9% | — |
| Kagi LLM Benchmark | 37.9% | — |
| ARC-AGI-1 | 63.7% | — |
| Thematic Generalization | 61.7% | — |
| EBR-Bench | — | 2.4% |
| ForecastBench | 57 | — |
Math Kimi K2.6 leads
GPT-5.4 mini: 45.5 (#75), Kimi K2.6: 57.0 (#41)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | 57.2% |
| FrontierMath Tier 4 | 9.8% | 25.6% |
| OTIS Mock AIME 2024-2025 | 88.9% | 96.1% |
| ProofBench | 21% | 16% |
| LMArena Math | 1419 | 1475 |
| FrontierMath (Feb 2025 set) | 28.3% | 39% |
| FrontierMath Tier 4 (v1) | 2.1% | 14.6% |
| MathArena Final-Answer Competitions | — | 72.9% |
Knowledge Kimi K2.6 leads
GPT-5.4 mini: 51.5 (#67), Kimi K2.6: 54.0 (#54)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| GPQA Diamond | 86.9% | 90.8% |
| SimpleQA Verified | 29.4% | 34.9% |
| Vectara Hallucination Rate | 5.5% | 10.8% |
| LMArena Expert | 1435 | 1491 |
Multimodal GPT-5.4 mini leads
GPT-5.4 mini: 39.7 (#56), Kimi K2.6: 31.6 (#103)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena Vision | 1245 | 1283 |
| Blueprint-Bench 2 | — | 3.9% |
| Furniture Assembly | — | 21.7% |
| LMArena Document | — | 1451 |
Multilingual Kimi K2.6 leads
GPT-5.4 mini: 51.9 (#96), Kimi K2.6: 54.9 (#37)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena Non-English | 1405 | 1446 |
| LMArena Chinese | 1446 | 1521 |
| LMArena French | 1440 | 1471 |
| LMArena German | 1409 | 1450 |
| LMArena Japanese | 1374 | 1443 |
| LMArena Korean | 1368 | 1427 |
| LMArena Russian | 1417 | 1446 |
| LMArena Spanish | 1405 | 1464 |
Instruction Following Kimi K2.6 leads
GPT-5.4 mini: 74.1 (#102), Kimi K2.6: 76.3 (#43)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena Instruction Following | 1405 | 1451 |
Long Context Kimi K2.6 leads
GPT-5.4 mini: 43.0 (#112), Kimi K2.6: 44.9 (#52)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena Longer Query | 1407 | 1468 |
Writing & Preference Kimi K2.6 leads
GPT-5.4 mini: 64.0 (#58), Kimi K2.6: 68.5 (#26)
| Benchmark | GPT-5.4 mini | Kimi K2.6 |
|---|---|---|
| LMArena Text | 1412 | 1455 |
| LMArena Creative Writing | 1370 | 1434 |
| EQ-Bench Creative Writing | 1665 | 1725 |
| LMArena Multi-Turn | 1429 | 1453 |
| EQ-Bench 4 | — | 1202 |
Frequently asked questions
Is GPT-5.4 mini better than Kimi K2.6?
Kimi K2.6 is the stronger model overall, scoring 47.7 to 45.0 on the Noometry Index.
Which is cheaper, GPT-5.4 mini or Kimi K2.6?
GPT-5.4 mini is cheaper. It lists at $0.75 per million input tokens and $4.50 per million output tokens; Kimi K2.6 lists at $0.95 and $4.
Is GPT-5.4 mini or Kimi K2.6 better for coding?
Kimi K2.6 scores higher on coding benchmarks: 50.7 versus 45.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 mini does, with 400K tokens against 262K.
How many benchmarks do GPT-5.4 mini and Kimi K2.6 share?
39 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Kimi K2.6 has 51.