GoogleGemma 4 E4B
Best configuration ranks #391 of 610 on the BenchLeader Index at 45.4 ±4.9. Released 2 Apr 2026.
- Blended price
- $0.040/M
- $0.020 in · $0.100 out
- Output speed
- 41 tok/s
- First answer
- 49 s
- Context
- 131k
- Overall index45
- Agents & tools41
- Knowledge55
- Instruction following46
- Multimodal35
- Long context41
- Composite40
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Default” means the publisher did not say which setting was used.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 43.1 | #443 | 42 tok/s | 0.74 s | $0.0000 | Agents & tools 40 · Composite 38 · Instruction following 43 · Knowledge 45 · Long context 37 · Multimodal 35 |
| defaultbest | 45.4 | #391 | 41 tok/s | 49 s | $0.0000 | Agents & tools 41 · Composite 40 · Instruction following 46 · Knowledge 55 · Long context 41 · Multimodal 35 |
Benchmark results
One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| GPQA Diamond (AA)not in index | 55.0%#386 | 57.6%#371 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 4.8%#392 | 3.8%#486 | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| Terminal-Bench Hard | 7.6%#232 | 8.3%#226 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 26.0%#277 | 20.8%#315 | Artificial Analysis |
Knowledge
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| AA-Omniscience | -41.0#258 | -19.7#182 | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| IFBench | 40.5%#247 | 44.2%#211 | Artificial Analysis |
Multimodal
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| MMMU-Pro | 51.2%#203 | 51.4%#202 | Artificial Analysis |
| MMMU-Pro (official)not in index | – | 52.6%#14 | MMMU |
Long context
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| AA-LCR | 24.0%#344 | 32.0%#315 | Artificial Analysis |
Composite
| Benchmark | no reasoning | default | Source |
|---|---|---|---|
| AA Intelligence Index | 7.5#394 | 8.9#347 | Artificial Analysis |
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0000 | – | 56.5 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0003 | – | 1.1 min |
| Code edit | 6,000 / 1,500 | $0.0003 | – | 1.4 min |
| Agentic coding session | 60,000 / 4,000 | $0.0016 | – | 2.4 min |
| Structured extraction | 2,000 / 200 | $0.0001 | – | 54.1 s |
See also
Data as of 9 Sept 2026. Compare these configurations.