NVIDIAGemma 3 12B IT
Best configuration ranks #564 of 610 on the BenchLeader Index at 37.2 ±5.3. Last measured 2 Sept 2026. Released 12 Mar 2025.
- Blended price
- $0.063/M
- $0.050 in · $0.100 out
- Output speed
- 5 tok/s
- First answer
- 1.17 s
- Context
- 131k
- Overall index37
- Reasoning37
- Coding28
- Agents & tools40
- Maths31
- Knowledge28
- Instruction following40
- Human preference51
- Multimodal28
- Long context29
- Composite34
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 | default | Source |
|---|---|---|
| GPQA Diamond | 39.5%#232 | Epoch AI Benchmarking Hub |
| LMArena Hard Prompts | 1332#207 | LMArena |
| GPQA Diamond (AA)not in index | 35.0%#466 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 4.2%#444 | Artificial Analysis |
Coding
| Benchmark | default | Source |
|---|---|---|
| SciCode | 17.4%#154 | SciCode |
| LMArena Coding | 1316#239 | LMArena |
| SciCode (AA)not in index | 16.4%#160 | Artificial Analysis |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| Terminal-Bench Hard | 0.8%#332 | Artificial Analysis |
| τ²-Bench Telecom (AA)not in index | 10.8%#364 | Artificial Analysis |
| BFCL Overall | 30.4%#49 | Berkeley Function Calling Leaderboard |
Maths
| Benchmark | default | Source |
|---|---|---|
| OTIS Mock AIME | 16.7%#205 | Epoch AI Benchmarking Hub |
Knowledge
| Benchmark | default | Source |
|---|---|---|
| AA-Omniscience | -77.1#452 | Artificial Analysis |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| IFBench | 36.7%#282 | Artificial Analysis |
Human preference
| Benchmark | default | Source |
|---|---|---|
| LMArena Text | 1342#189 | LMArena |
Multimodal
| Benchmark | default | Source |
|---|---|---|
| MMMU-Pro | 37.5%#235 | Artificial Analysis |
| MMMU (validation) | 59.6%#32 | MMMU |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 8.3%#400 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| Epoch Capabilities Indexnot in index | 123.5#154 | Epoch AI Benchmarking Hub |
| AA Intelligence Index | 3.8#539 | Artificial Analysis |
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| DeepInfra | 5 tok/s | 1.17 s | $0.050 | $0.150 | 131k | bf16 |
Price history
Listed price per 1M tokens over time, as recorded by OpenRouter for the provider with the longest history.
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 | $0.0001 | – | 1.0 min |
| Summarise a 30-page report | 12,000 / 600 | $0.0007 | – | 2.0 min |
| Code edit | 6,000 / 1,500 | $0.0005 | – | 5.0 min |
| Agentic coding session | 60,000 / 4,000 | $0.0034 | – | 13.4 min |
| Structured extraction | 2,000 / 200 | $0.0001 | – | 41.2 s |
See also
Data as of 9 Sept 2026. Compare with another model.