MetaLlama 3.2 3B
Best configuration ranks #573 of 610 on the BenchLeader Index at 36.7 ±2.4. Last measured 2 Sept 2026. Released 18 Sept 2024.
- Blended price
- $0.120/M
- $0.050 in · $0.330 out
- Output speed
- 28 tok/s
- First answer
- 0.54 s
- Context
- 128k
- Overall index37
- Reasoning30
- Coding28
- Agents & tools35
- Instruction following31
- Human preference30
- Long context27
- Composite36
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 |
|---|---|---|
| LMArena Hard Prompts | 1168#305 | LMArena |
| GPQA Diamond (AA)not in index | 25.4%#511 | Artificial Analysis |
| Humanity's Last Exam (AA)not in index | 5.3%#355 | Artificial Analysis |
Coding
| Benchmark | default | Source |
|---|---|---|
| LMArena Coding | 1176#312 | LMArena |
Agents & tools
| Benchmark | default | Source |
|---|---|---|
| τ²-Bench Telecom (AA)not in index | 21.1%#312 | Artificial Analysis |
| BFCL Overall | 21.9%#73 | Berkeley Function Calling Leaderboard |
Instruction following
| Benchmark | default | Source |
|---|---|---|
| IFBench | 26.2%#369 | Artificial Analysis |
Human preference
| Benchmark | default | Source |
|---|---|---|
| LMArena Text | 1166#306 | LMArena |
Long context
| Benchmark | default | Source |
|---|---|---|
| AA-LCR | 4.3%#416 | Artificial Analysis |
Composite
| Benchmark | default | Source |
|---|---|---|
| AA Intelligence Index | 5.7#476 | 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 |
|---|---|---|---|---|---|---|
| Cloudflare | 33 tok/s | 0.35 s | $0.051 | $0.335 | 80k | – |
| Parasail | 23 tok/s | 0.72 s | $0.050 | $0.330 | 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 | – | 11.3 s |
| Summarise a 30-page report | 12,000 / 600 | $0.0008 | – | 22.0 s |
| Code edit | 6,000 / 1,500 | $0.0008 | – | 54.1 s |
| Agentic coding session | 60,000 / 4,000 | $0.0043 | – | 2.4 min |
| Structured extraction | 2,000 / 200 | $0.0002 | – | 7.7 s |
See also
Data as of 9 Sept 2026. Compare with another model.