Model comparison
Claude Haiku 4.5 vs Gemma 4 31B IT
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 39.5 on the Noometry Index.
Last verified . 30 shared benchmarks.
GoogleSummary
- They share 30 benchmarks with published results for both. Claude Haiku 4.5 scores higher in 2 categories and Gemma 4 31B IT in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in multimodal, where Gemma 4 31B IT leads 41.6 to 26.8.
- The biggest single-benchmark swing is NYT Connections (extended): 14.3% for Claude Haiku 4.5 and 70.6% for Gemma 4 31B IT.
- Gemma 4 31B IT is cheaper at $0.09 / $0.34 per million input/output tokens, against $1 / $5 for Claude Haiku 4.5.
- Gemma 4 31B IT accepts more context: 262K tokens versus 200K.
- Gemma 4 31B IT has downloadable open weights; the other is API-only.
Side by side
| Claude Haiku 4.5 | Gemma 4 31B IT | |
|---|---|---|
| Provider | Anthropic | |
| Noometry Index | 39.5 | 43.5 |
| Released | 2025-10-15 | 2026-04-02 |
| Weights | Proprietary | Open |
| Context window | 200K | 262K |
| Max output | 64K | 33K |
| Input $ / M tokens | $1 | $0.09 |
| Output $ / M tokens | $5 | $0.34 |
| Results tracked | 53 | 35 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude Haiku 4.5 leads
Claude Haiku 4.5: 44.0 (#78), Gemma 4 31B IT: 42.3 (#108)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena WebDev | 1330 | 1366 |
| SciCode | 43.3% | 43.4% |
| WeirdML | 45.4% | 52.3% |
| LMArena Coding | 1453 | 1459 |
| ALE-Bench | 653.48 | 925.5 |
| SWE-bench Verified (bash only) | 66.6% | — |
| SWE-bench Multilingual | 64.7% | — |
Agentic & Tool Use Not comparable
Claude Haiku 4.5: 33.6 (#52), Gemma 4 31B IT: —
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| Terminal-Bench | 35.5% | — |
| Berkeley Function Calling Leaderboard | 68.7% | — |
| DeepResearch Bench | 45.5% | — |
| BALROG | 31.2% | — |
| ExploitBench | 13.7% | — |
| Vending-Bench 2 | 458.89 | — |
Reasoning Gemma 4 31B IT leads
Claude Haiku 4.5: 15.1 (#320), Gemma 4 31B IT: 27.2 (#122)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| NYT Connections (extended) | 14.3% | 70.6% |
| CritPt | 0% | 1.4% |
| Chess Puzzles | 8% | 5% |
| LMArena Hard Prompts | 1420 | 1448 |
| DTBench | 73.6% | 82.7% |
| LMCA | 30.9% | 39.3% |
| Epoch Capabilities Index | 142.41 | 142.74 |
| ARC-AGI-2 | 4% | — |
| Kagi LLM Benchmark | — | 63.5% |
| ARC-AGI-1 | 47.7% | — |
| Thematic Generalization | — | 53% |
| Surface Evolver Bench | — | 30.6% |
| ForecastBench | 61.4 | — |
Math Claude Haiku 4.5 leads
Claude Haiku 4.5: 44.9 (#78), Gemma 4 31B IT: 43.2 (#81)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 66.7% | 73.3% |
| LMArena Math | 1396 | 1465 |
| Omni-MATH | 56.1% | — |
| MATH Level 5 | 96.4% | — |
| FrontierMath (Feb 2025 set) | 5.9% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Too close to call
Claude Haiku 4.5: 37.7 (#153), Gemma 4 31B IT: 37.9 (#151)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| GPQA Diamond | 71.2% | 75.8% |
| SimpleQA Verified | 13.2% | 10.4% |
| Vectara Hallucination Rate | 9.8% | 7.4% |
| LMArena Expert | 1442 | 1465 |
| MMLU-Pro | 77.7% | — |
| GPQA (HELM) | 60.5% | — |
Multimodal Gemma 4 31B IT leads
Claude Haiku 4.5: 26.8 (#118), Gemma 4 31B IT: 41.6 (#34)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena Document | 1420 | 1425 |
| LMArena Vision | — | 1277 |
| Blueprint-Bench 2 | 0% | — |
Multilingual Gemma 4 31B IT leads
Claude Haiku 4.5: 49.9 (#129), Gemma 4 31B IT: 53.8 (#57)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena Non-English | 1377 | 1431 |
| LMArena Chinese | 1417 | 1476 |
| LMArena French | 1408 | 1435 |
| LMArena Russian | 1381 | 1460 |
| LMArena Spanish | 1420 | 1444 |
| LMArena German | 1375 | — |
| LMArena Japanese | 1339 | — |
| LMArena Korean | 1347 | — |
Instruction Following Gemma 4 31B IT leads
Claude Haiku 4.5: 71.4 (#149), Gemma 4 31B IT: 75.5 (#61)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena Instruction Following | 1414 | 1433 |
| IFEval | 80.1% | — |
Long Context Too close to call
Claude Haiku 4.5: 43.6 (#92), Gemma 4 31B IT: 44.2 (#71)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena Longer Query | 1427 | 1446 |
Writing & Preference Gemma 4 31B IT leads
Claude Haiku 4.5: 57.9 (#123), Gemma 4 31B IT: 60.5 (#96)
| Benchmark | Claude Haiku 4.5 | Gemma 4 31B IT |
|---|---|---|
| LMArena Text | 1396 | 1443 |
| LMArena Creative Writing | 1372 | 1415 |
| EQ-Bench 4 | 1064 | 1120 |
| LMArena Multi-Turn | 1409 | 1452 |
| EQ-Bench Creative Writing | — | 1368 |
| WildBench | 83.9% | — |
Frequently asked questions
Is Claude Haiku 4.5 better than Gemma 4 31B IT?
Gemma 4 31B IT is the stronger model overall, scoring 43.5 to 39.5 on the Noometry Index.
Which is cheaper, Claude Haiku 4.5 or Gemma 4 31B IT?
Gemma 4 31B IT is cheaper. It lists at $0.09 per million input tokens and $0.34 per million output tokens; Claude Haiku 4.5 lists at $1 and $5.
Is Claude Haiku 4.5 or Gemma 4 31B IT better for coding?
Claude Haiku 4.5 scores higher on coding benchmarks: 44.0 versus 42.3 in the Noometry coding category.
Which has the bigger context window?
Gemma 4 31B IT does, with 262K tokens against 200K.
How many benchmarks do Claude Haiku 4.5 and Gemma 4 31B IT share?
30 benchmarks have published results for both models. Claude Haiku 4.5 has 53 scored results on Noometry and Gemma 4 31B IT has 35.