granite-4.1-3b: benchmark scores, price and ranking

granite-4.1-3b from ibm-granite ranks #639 of 743 LLMs on the LLMs Tiger aggregate benchmark score (44.4), based on 3 benchmark results captured up to 2026-10-13.

Organisation: ibm-granite
Type: Open-weight / local model
Open weights: Yes
Aggregate rank: #639 of 743 (score 44.4)

Relative to other models, granite-4.1-3b is strongest on GSM8K (#9 of 17) and weakest on GPQA Diamond (#245 of 275).

Across the 3 benchmarks where at least ten models are scored, granite-4.1-3b is in the top 10% on 0 and in the top half on 0.

granite-4.1-3b benchmark results

granite-4.1-3b benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
GSM8K	86.9%	#9 of 17	HuggingFace GSM8K leaderboard	2026-10-13
MMLU-Pro	49.8%	#78 of 136	HuggingFace MMLU-Pro leaderboard	2026-10-13
GPQA Diamond	31.7%	#245 of 275	HuggingFace GPQA Diamond leaderboard	2026-10-13

About granite-4.1-3b

granite-4.1-3b is an open-weight model that can run on your own hardware from ibm-granite. No API price is listed for it yet. We hold 3 benchmark results for it across 2 areas, and none is in the top tenth of its board.

Where granite-4.1-3b stands

On our aggregate score, granite-4.1-3b is #639 of 743 ranked models, in the lower half. Its score is 44.4; the scale is centred on 50, so it sits 5.6 points below the typical model. That beats 14% of the ranked field. Among the 9 ranked ibm-granite models it is #5. The next one up is granite-3.0-8b-base at 44.6. The next one down is granite-3.1-2b-base at 43.6.

granite-4.1-3b, benchmark by benchmark

granite-4.1-3b lands below the middle on GSM8K: 86.9%, #9 of 17, with 47% of the field scoring higher. mimo-v2.5-pro leads at 99.6%; the closest result above is DeepSeek-V3 at 89.3%. It covers grade-school math word problems. The median score on this board is also 86.9%. Source: HuggingFace GSM8K leaderboard, captured 2026-10-13.

granite-4.1-3b lands below the middle on MMLU-Pro: 49.8%, #78 of 136, with 57% of the field scoring higher. Darwin-180B-RSI leads at 88.1%; the closest result above is Qwen1.5-110B at 49.9%. It covers 57-subject knowledge + reasoning. The board median is 55.3%, so granite-4.1-3b is 5.4 points below it. Source: HuggingFace MMLU-Pro leaderboard, captured 2026-10-13.

GPQA Diamond is a weak spot. granite-4.1-3b scores 31.7%, #245 of 275, behind 89% of the models measured. The leader, GPT-6 Astra (max), reaches 95.8%. It covers expert-level science questions. The board median is 71.7%, so granite-4.1-3b is 40.0 points below it. Source: HuggingFace GPQA Diamond leaderboard, captured 2026-10-13.

Strengths and weaknesses

Reasoning, maths and science: weak, ahead of 27% of the field on average across 2 benchmarks. Its best board here is MMLU-Pro (#78 of 136) and its weakest is GPQA Diamond (#245 of 275).

Reading these numbers

The 3 scores on this page come from 3 sources: HuggingFace GPQA Diamond leaderboard (1), HuggingFace GSM8K leaderboard (1), HuggingFace MMLU-Pro leaderboard (1). No official leaderboard supplied any of them, so treat them as third-party measurements. With 3 results the aggregate rank can move a lot when one new score arrives.

Models ranked nearby

#636 DeepSeek-V3.2 (Thinking; Novita) (44.5)
#637 mistral-small-3.1-2503 (44.4)
#638 Yi-1.5-34B-Chat (44.4)
#640 Laguna M.1 (44.4)
#641 grok-code-fast-1 (44.3)
#642 Qwen1.5-7B-Chat (44.3)