IBMGranite 4.2 3B
Granite 4.2 3B is an IBM open-weights reasoning model, released 25 Aug 2026. Its best configuration ranks #519 of 775 on the BenchLeader Index at 44.8 ±4.6, in the lower half. The ± is the point: 301 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in knowledge (55) and lowest in agents & tools (33). At $0.053 per million tokens blended it is among the cheapest fifth of ranked models. Output speed of 51 tokens per second puts it slower than most, with a first answer in 39.7 s. Last measured 15 Oct 2026.
- Blended price
- $0.053/M
- $0.030 in · $0.120 out
- Output speed
- 51 tok/s
- measured by Artificial Analysis
- First answer
- 40 s
- Context
- 131k
- Full answer
- 49 s
- median, reasoning included
- Cost per run
- $0.0060
- one full Intelligence Index run
- Released
- 25 Aug 2026
- Overall index45
- Reasoning43
- Coding50
- Agents & tools33
- Knowledge55
- Human preference44
- Long context36
- Composite40
Versions
IBM has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Granite 4.2 3Bthis page | 25 Aug 2026 | 44.8 | #519 |
| Granite 4.1 3B | 29 Apr 2026 | 39.2 | #682 |
How its index has moved
9 Sept 2026 to 20 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. 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.
Metadata
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA Openness Indexnot in index | 72.2#20 | Artificial Analysis |
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Hard Prompts | 1311#257 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 55.9%#397 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 6.6%#379 | Artificial Analysis | |
| CritPt | 0.0%#310 | Artificial Analysis |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Coding | 1379#217 | Arena (LMArena) | |
| SciCode (AA)not in index | 25.4%#216 | Artificial Analysis | |
| Terminal-Bench 4.0 (AA)not in index | 0.0%#149 | Artificial Analysis | |
| Terminal-Bench 2.1 (AA)not in index | 13.9%#182 | Artificial Analysis |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| GDPval-AA v2.1not in index | 0.0%#218 | Artificial Analysis | |
| τ³-Banking (AA) | 5.6%#173 | Artificial Analysis | |
| AutomationBenchnot in index | 0.4%#208 | Artificial Analysis | |
| GDP.pdf (AA)not in index | 0.2%#215 | Artificial Analysis | |
| AA-Briefcase v1.1not in index | 98#58 | Artificial Analysis |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -14.7#219 | Artificial Analysis | |
| AA-Omniscience: accuracynot in index | 9.2%#501 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 73.7%#31 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Instruction Followingnot in index | 1284#258 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1290#268 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1206#297 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1286#260 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1301#262 | Arena (LMArena) |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 24.3%#409 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA Intelligence Index v4.3.2 | 9.1#405 | 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 | – | 45.6 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0004 | – | 51.4 s |
| Code edit | 6,000 / 1,500 | $0.0004 | – | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.0023 | – | 2.0 min |
| Structured extraction | 2,000 / 200 | $0.0001 | – | 43.6 s |
More from IBM
See also
Data as of 21 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Granite 4.2 3B: benchmarks, pricing, speed and rank”, https://www.benchleader.com/models/granite-4-2-3b, data as of 21 Oct 2026.