Model comparison
GPT-5.4 mini vs Kimi K2.5
Kimi K2.5 is the stronger model overall, scoring 48.1 to 45.0 on the Noometry Index.
Last verified . 37 shared benchmarks.
Summary
- They share 37 benchmarks with published results for both. GPT-5.4 mini scores higher in 0 categories and Kimi K2.5 in 10 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in long context, where Kimi K2.5 leads 52.1 to 43.0.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 37.9% for GPT-5.4 mini and 78.5% for Kimi K2.5.
- Kimi K2.5 is cheaper at $0.45 / $2.25 per million input/output tokens, against $0.75 / $4.50 for GPT-5.4 mini.
- GPT-5.4 mini accepts more context: 400K tokens versus 262K.
- Kimi K2.5 has downloadable open weights; the other is API-only.
Side by side
| GPT-5.4 mini | Kimi K2.5 | |
|---|---|---|
| Provider | OpenAI | Moonshot AI |
| Noometry Index | 45.0 | 48.1 |
| Released | 2026-03-17 | 2026-01-27 |
| Weights | Proprietary | Open |
| Context window | 400K | 262K |
| Max output | 128K | 262K |
| Input $ / M tokens | $0.75 | $0.45 |
| Output $ / M tokens | $4.50 | $2.25 |
| Results tracked | 46 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
GPT-5.4 mini: 45.2 (#72), Kimi K2.5: 48.8 (#53)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena WebDev | 1397 | 1437 |
| SciCode | 49.9% | 49% |
| WeirdML | 60.3% | 45.6% |
| LMArena Coding | 1438 | 1474 |
| ALE-Bench | 1,189 | 821.65 |
| SWE-bench Verified | — | 73.8% |
| FrontierCode | 27% | — |
| SWE-bench Verified (bash only) | — | 70.8% |
| SWE-bench Multilingual | — | 67.3% |
Agentic & Tool Use Kimi K2.5 leads
GPT-5.4 mini: 29.9 (#81), Kimi K2.5: 34.2 (#48)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | — | 43.2% |
| DeepResearch Bench | 36.3% | — |
| OSWorld | — | 63.3% |
| Vending-Bench 2 | — | 1,198 |
Reasoning Too close to call
GPT-5.4 mini: 30.4 (#85), Kimi K2.5: 31.2 (#80)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| ARC-AGI-2 | 18.9% | 11.8% |
| Kagi LLM Benchmark | 37.9% | 78.5% |
| NYT Connections (extended) | 61.8% | 69.9% |
| ARC-AGI-1 | 63.7% | 65.3% |
| CritPt | 10% | 3.1% |
| Chess Puzzles | 24% | 12% |
| Thematic Generalization | 61.7% | 69.4% |
| LMArena Hard Prompts | 1424 | 1453 |
| Epoch Capabilities Index | 148.84 | 148.03 |
| SimpleBench | — | 46.8% |
| EnigmaEval | — | 3.4% |
| Mystery Game Puzzles | 11% | — |
| DTBench | 80% | — |
| LMCA | 40.8% | — |
| ForecastBench | 57 | — |
Math Kimi K2.5 leads
GPT-5.4 mini: 45.5 (#75), Kimi K2.5: 51.8 (#53)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 88.9% | 92.2% |
| LMArena Math | 1419 | 1470 |
| FrontierMath (Feb 2025 set) | 28.3% | 27.9% |
| FrontierMath Tier 4 (v1) | 2.1% | 4.2% |
| FrontierMath (Tiers 1-3) | 51.2% | — |
| FrontierMath Tier 4 | 9.8% | — |
| MathArena Final-Answer Competitions | — | 62.3% |
| ProofBench | 21% | — |
Knowledge Kimi K2.5 leads
GPT-5.4 mini: 51.5 (#67), Kimi K2.5: 53.6 (#56)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | 86.9% | 87.6% |
| SimpleQA Verified | 29.4% | 34.3% |
| Vectara Hallucination Rate | 5.5% | 14.2% |
| LMArena Expert | 1435 | 1466 |
| Humanity's Last Exam | — | 24.4% |
Multimodal Kimi K2.5 leads
GPT-5.4 mini: 39.7 (#56), Kimi K2.5: 41.1 (#39)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena Vision | 1245 | 1269 |
| LMArena Document | — | 1430 |
Multilingual Kimi K2.5 leads
GPT-5.4 mini: 51.9 (#96), Kimi K2.5: 53.9 (#53)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1405 | 1433 |
| LMArena Chinese | 1446 | 1495 |
| LMArena French | 1440 | 1454 |
| LMArena German | 1409 | 1441 |
| LMArena Japanese | 1374 | 1421 |
| LMArena Korean | 1368 | 1410 |
| LMArena Russian | 1417 | 1435 |
| LMArena Spanish | 1405 | 1450 |
Instruction Following Kimi K2.5 leads
GPT-5.4 mini: 74.1 (#102), Kimi K2.5: 75.3 (#64)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1405 | 1431 |
Long Context Kimi K2.5 leads
GPT-5.4 mini: 43.0 (#112), Kimi K2.5: 52.1 (#7)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena Longer Query | 1407 | 1445 |
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |
Writing & Preference Kimi K2.5 leads
GPT-5.4 mini: 64.0 (#58), Kimi K2.5: 65.1 (#53)
| Benchmark | GPT-5.4 mini | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1412 | 1445 |
| LMArena Creative Writing | 1370 | 1423 |
| EQ-Bench Creative Writing | 1665 | 1579 |
| LMArena Multi-Turn | 1429 | 1444 |
Frequently asked questions
Is GPT-5.4 mini better than Kimi K2.5?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 45.0 on the Noometry Index.
Which is cheaper, GPT-5.4 mini or Kimi K2.5?
Kimi K2.5 is cheaper. It lists at $0.45 per million input tokens and $2.25 per million output tokens; GPT-5.4 mini lists at $0.75 and $4.50.
Is GPT-5.4 mini or Kimi K2.5 better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.8 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 K2.5 share?
37 benchmarks have published results for both models. GPT-5.4 mini has 46 scored results on Noometry and Kimi K2.5 has 51.