Model comparison
Gemma 4 31B IT vs Kimi K2.5
Kimi K2.5 is the stronger model overall, scoring 48.1 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 5.9× less per token, which makes it the better buy when Kimi K2.5's lead doesn't matter for your workload.
Last verified . 31 shared benchmarks.
GoogleSummary
- They share 31 benchmarks with published results for both. Gemma 4 31B IT scores higher in 2 categories and Kimi K2.5 in 7 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Kimi K2.5 leads 53.6 to 37.9.
- The biggest single-benchmark swing is SimpleQA Verified: 10.4% for Gemma 4 31B IT and 34.3% for Kimi K2.5.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $0.45 / $2.25 for Kimi K2.5.
Side by side
| Gemma 4 31B IT | Kimi K2.5 | |
|---|---|---|
| Provider | Moonshot AI | |
| Noometry Index | 43.5 | 48.1 |
| Released | 2026-04-02 | 2026-01-27 |
| Weights | Open | Open |
| Context window | 262K | 262K |
| Max output | 33K | 262K |
| Input $ / M tokens | $0.09 | $0.45 |
| Output $ / M tokens | $0.34 | $2.25 |
| Results tracked | 35 | 51 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Kimi K2.5 leads
Gemma 4 31B IT: 42.3 (#108), Kimi K2.5: 48.8 (#53)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena WebDev | 1366 | 1437 |
| SciCode | 43.4% | 49% |
| WeirdML | 52.3% | 45.6% |
| LMArena Coding | 1459 | 1474 |
| ALE-Bench | 925.5 | 821.65 |
| SWE-bench Verified | — | 73.8% |
| SWE-bench Verified (bash only) | — | 70.8% |
| SWE-bench Multilingual | — | 67.3% |
Agentic & Tool Use Not comparable
Gemma 4 31B IT: —, Kimi K2.5: 34.2 (#48)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| Terminal-Bench | — | 43.2% |
| OSWorld | — | 63.3% |
| Vending-Bench 2 | — | 1,198 |
Reasoning Kimi K2.5 leads
Gemma 4 31B IT: 27.2 (#122), Kimi K2.5: 31.2 (#80)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| Kagi LLM Benchmark | 63.5% | 78.5% |
| NYT Connections (extended) | 70.6% | 69.9% |
| CritPt | 1.4% | 3.1% |
| Chess Puzzles | 5% | 12% |
| Thematic Generalization | 53% | 69.4% |
| LMArena Hard Prompts | 1448 | 1453 |
| Epoch Capabilities Index | 142.74 | 148.03 |
| ARC-AGI-2 | — | 11.8% |
| SimpleBench | — | 46.8% |
| ARC-AGI-1 | — | 65.3% |
| EnigmaEval | — | 3.4% |
| DTBench | 82.7% | — |
| LMCA | 39.3% | — |
| Surface Evolver Bench | 30.6% | — |
Math Kimi K2.5 leads
Gemma 4 31B IT: 43.2 (#81), Kimi K2.5: 51.8 (#53)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 73.3% | 92.2% |
| LMArena Math | 1465 | 1470 |
| MathArena Final-Answer Competitions | — | 62.3% |
| FrontierMath (Feb 2025 set) | — | 27.9% |
| FrontierMath Tier 4 (v1) | — | 4.2% |
Knowledge Kimi K2.5 leads
Gemma 4 31B IT: 37.9 (#151), Kimi K2.5: 53.6 (#56)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| GPQA Diamond | 75.8% | 87.6% |
| SimpleQA Verified | 10.4% | 34.3% |
| Vectara Hallucination Rate | 7.4% | 14.2% |
| LMArena Expert | 1465 | 1466 |
| Humanity's Last Exam | — | 24.4% |
Multimodal Too close to call
Gemma 4 31B IT: 41.6 (#34), Kimi K2.5: 41.1 (#39)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena Vision | 1277 | 1269 |
| LMArena Document | 1425 | 1430 |
Multilingual Too close to call
Gemma 4 31B IT: 53.8 (#57), Kimi K2.5: 53.9 (#53)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena Non-English | 1431 | 1433 |
| LMArena Chinese | 1476 | 1495 |
| LMArena French | 1435 | 1454 |
| LMArena Russian | 1460 | 1435 |
| LMArena Spanish | 1444 | 1450 |
| LMArena German | — | 1441 |
| LMArena Japanese | — | 1421 |
| LMArena Korean | — | 1410 |
Instruction Following Too close to call
Gemma 4 31B IT: 75.5 (#61), Kimi K2.5: 75.3 (#64)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena Instruction Following | 1433 | 1431 |
Long Context Kimi K2.5 leads
Gemma 4 31B IT: 44.2 (#71), Kimi K2.5: 52.1 (#7)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena Longer Query | 1446 | 1445 |
| Fiction.LiveBench | — | 86.1% |
| CL-bench | — | 19.3% |
| CL-bench Life | — | 13.2% |
Writing & Preference Kimi K2.5 leads
Gemma 4 31B IT: 60.5 (#96), Kimi K2.5: 65.1 (#53)
| Benchmark | Gemma 4 31B IT | Kimi K2.5 |
|---|---|---|
| LMArena Text | 1443 | 1445 |
| LMArena Creative Writing | 1415 | 1423 |
| EQ-Bench Creative Writing | 1368 | 1579 |
| LMArena Multi-Turn | 1452 | 1444 |
| EQ-Bench 4 | 1120 | — |
Frequently asked questions
Is Gemma 4 31B IT better than Kimi K2.5?
Kimi K2.5 is the stronger model overall, scoring 48.1 to 43.5 on the Noometry Index. Gemma 4 31B IT costs 5.9× less per token, which makes it the better buy when Kimi K2.5's lead doesn't matter for your workload.
Which is cheaper, Gemma 4 31B IT or Kimi K2.5?
Gemma 4 31B IT is cheaper. It lists at $0.09 per million input tokens and $0.34 per million output tokens; Kimi K2.5 lists at $0.45 and $2.25.
Is Gemma 4 31B IT or Kimi K2.5 better for coding?
Kimi K2.5 scores higher on coding benchmarks: 48.8 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
Both accept 262K tokens.
How many benchmarks do Gemma 4 31B IT and Kimi K2.5 share?
31 benchmarks have published results for both models. Gemma 4 31B IT has 35 scored results on Noometry and Kimi K2.5 has 51.