Model comparison
GPT-5.4 mini vs Kimi K3
Kimi K3 is the stronger model overall, scoring 59.5 to 45.0 on the Noometry Index. GPT-5.4 mini costs 3.6× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Last verified . 39 shared benchmarks.
Summary
- They share 39 benchmarks with published results for both. GPT-5.4 mini scores higher in 1 category and Kimi K3 in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Kimi K3 leads 63.0 to 30.4.
- The biggest single-benchmark swing is ProofBench: 21% for GPT-5.4 mini and 87% for Kimi K3.
- GPT-5.4 mini is cheaper at $0.75 / $4.50 per million input/output tokens, against $3 / $15 for Kimi K3.
- Kimi K3 accepts more context: 1.05M tokens versus 400K.
- Kimi K3 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 mini | Kimi K3 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 45.0 | 59.5 |
| Released | 2026-03-17 | 2026-07-16 |
| Weights | Proprietary | Open |
| Context window | 400K | 1.05M |
| Max output | 128K | 1.05M |
| Input $ / M tokens | $0.75 | $3 |
| Output $ / M tokens | $4.50 | $15 |
| Results tracked | 46 | 53 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K3 leads
GPT-5.4 mini: 45.2 (#72), Kimi K3: 61.0 (#10)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| FrontierCode | 27% | 44.2% |
| LMArena WebDev | 1397 | 1654 |
| SciCode | 49.9% | 59.5% |
| WeirdML | 60.3% | 82.6% |
| LMArena Coding | 1438 | 1508 |
| ALE-Bench | 1,189 | 1,524 |
| DeepSWE | — | 68.5% |
| FrontierSWE | — | 25.9% |
Agentic & Tool Use Kimi K3 leads
GPT-5.4 mini: 29.9 (#81), Kimi K3: 41.8 (#20)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| APEX-Agents | — | 50.6% |
| τ²-bench Banking | — | 37.1% |
| DeepResearch Bench | 36.3% | — |
| PostTrainBench | — | 32% |
| GBAEval | — | 48.3% |
| GDP.pdf | — | 19% |
| Vending-Bench 2 | — | 5,165 |
Reasoning Kimi K3 leads
GPT-5.4 mini: 30.4 (#85), Kimi K3: 63.0 (#17)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| ARC-AGI-2 | 18.9% | 60.4% |
| NYT Connections (extended) | 61.8% | 93.6% |
| ARC-AGI-1 | 63.7% | 94.5% |
| CritPt | 10% | 23.4% |
| Chess Puzzles | 24% | 39% |
| LMArena Hard Prompts | 1424 | 1496 |
| Mystery Game Puzzles | 11% | 26% |
| DTBench | 80% | 91.2% |
| LMCA | 40.8% | 52.7% |
| Epoch Capabilities Index | 148.84 | 157.45 |
| ForecastBench | 57 | 61.1 |
| SimpleBench | — | 60.7% |
| Kagi LLM Benchmark | 37.9% | — |
| Thematic Generalization | 61.7% | — |
| Surface Evolver Bench | — | 95% |
Math Kimi K3 leads
GPT-5.4 mini: 45.5 (#75), Kimi K3: 74.2 (#16)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| FrontierMath (Tiers 1-3) | 51.2% | 72.2% |
| FrontierMath Tier 4 | 9.8% | 39% |
| OTIS Mock AIME 2024-2025 | 88.9% | 97.2% |
| ProofBench | 21% | 87% |
| LMArena Math | 1419 | 1491 |
| MathArena Final-Answer Competitions | — | 87.8% |
| FrontierMath (Feb 2025 set) | 28.3% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Kimi K3 leads
GPT-5.4 mini: 51.5 (#67), Kimi K3: 63.2 (#21)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| GPQA Diamond | 86.9% | 93.1% |
| SimpleQA Verified | 29.4% | 50.6% |
| LMArena Expert | 1435 | 1521 |
| Vectara Hallucination Rate | 5.5% | — |
Multimodal GPT-5.4 mini leads
GPT-5.4 mini: 39.7 (#56), Kimi K3: 37.8 (#70)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| LMArena Vision | 1245 | — |
| Blueprint-Bench 2 | — | 29.5% |
| Furniture Assembly | — | 34.2% |
Multilingual Kimi K3 leads
GPT-5.4 mini: 51.9 (#96), Kimi K3: 56.3 (#21)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| LMArena Non-English | 1405 | 1466 |
| LMArena Chinese | 1446 | 1529 |
| LMArena French | 1440 | 1491 |
| LMArena German | 1409 | 1488 |
| LMArena Japanese | 1374 | 1487 |
| LMArena Korean | 1368 | 1458 |
| LMArena Russian | 1417 | 1482 |
| LMArena Spanish | 1405 | 1472 |
Instruction Following Kimi K3 leads
GPT-5.4 mini: 74.1 (#102), Kimi K3: 77.7 (#14)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| LMArena Instruction Following | 1405 | 1483 |
Long Context Kimi K3 leads
GPT-5.4 mini: 43.0 (#112), Kimi K3: 45.8 (#29)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| LMArena Longer Query | 1407 | 1494 |
Writing & Preference Kimi K3 leads
GPT-5.4 mini: 64.0 (#58), Kimi K3: 76.6 (#4)
| Benchmark | GPT-5.4 mini | Kimi K3 |
|---|---|---|
| LMArena Text | 1412 | 1476 |
| LMArena Creative Writing | 1370 | 1454 |
| EQ-Bench Creative Writing | 1665 | 2082 |
| LMArena Multi-Turn | 1429 | 1488 |
| EQ-Bench 4 | — | 1339 |
Frequently asked questions
Is GPT-5.4 mini better than Kimi K3?
Kimi K3 is the stronger model overall, scoring 59.5 to 45.0 on the Noometry Index. GPT-5.4 mini costs 3.6× less per token, which makes it the better buy when Kimi K3's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 mini or Kimi K3?
GPT-5.4 mini is cheaper. It lists at $0.75 per million input tokens and $4.50 per million output tokens; Kimi K3 lists at $3 and $15.
Is GPT-5.4 mini or Kimi K3 better for coding?
Kimi K3 scores higher on coding benchmarks: 61.0 versus 45.2 in the Noometry coding category.
Which has the bigger context window?
Kimi K3 does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5.4 mini and Kimi K3 share?
39 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Kimi K3 has 53.