Mistral Small 4: benchmark scores, price and ranking

Mistral Small 4 from Mistral AI ranks #513 of 743 LLMs on the LLMs Tiger aggregate benchmark score (46.5), based on 3 benchmark results captured up to 2026-10-13.

Organisation: Mistral AI
Type: Open-weight / local model
Open weights: Yes
Released: 2026-03-16 (source: epoch)
Cheapest API price: $0.150 in / $0.600 out per million tokens (via openrouter, observed 2026-10-13T20:47Z)
Aggregate rank: #513 of 743 (score 46.5)

Relative to other models, Mistral Small 4 is strongest on DTBench (#71 of 146) and weakest on ALE-Bench (#59 of 79).

Across the 3 benchmarks where at least ten models are scored, Mistral Small 4 is in the top 10% on 0 and in the top half on 1.

Its cheapest listed input price, $0.150 per million tokens, is lower than that of 72% of the 427 models with a listed price.

Mistral Small 4 benchmark results

Mistral Small 4 benchmark scores, ranked against all models measured on each benchmark
Benchmark	Score	Rank	Source	Captured
DTBench	70.9%	#71 of 146	Epoch AI Benchmarking Hub	2026-10-13
LMCA	20.6%	#72 of 108	Epoch AI Benchmarking Hub	2026-10-13
ALE-Bench	498	#59 of 79	Epoch AI Benchmarking Hub	2026-10-13

About Mistral Small 4

Mistral Small 4 is an open-weight model that can run on your own hardware from Mistral AI. It was released on 2026-03-16, according to Epoch AI. The cheapest API listing we track charges $0.150 per million input tokens and $0.600 per million output tokens, through openrouter. We hold 3 benchmark results for it across 3 areas, and none is in the top tenth of its board.

Where Mistral Small 4 stands

On our aggregate score, Mistral Small 4 is #513 of 743 ranked models, in the lower half. Its score is 46.5; the scale is centred on 50, so it sits 3.5 points below the typical model. That beats 31% of the ranked field. Among the 44 ranked Mistral AI models it is #16. The next one up is mistral-small-2402 at 46.7. The next one down is ollama/mistral-small at 46.4.

Mistral Small 4, benchmark by benchmark

DTBench puts Mistral Small 4 at 70.9%, #71 of 146. That beats 51% of the models tested, but 28.0 points short of Claude Opus 5.5 (max) (98.9%). The results either side are Qwen3.5 9B (71.2%) and qwen3-30b-a3b-thinking-2507 (69.3%). It covers dynamic tool-use benchmark. The board median is 68.8%, so Mistral Small 4 is 2.1 points above it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Mistral Small 4 lands below the middle on LMCA: 20.6%, #72 of 108, with 66% of the field scoring higher. Claude Opus 5.5 (max) leads at 68.2%; the closest result above is GPT-4.1 mini at 21.1%. It covers language-model capability assessment. The board median is 29.2%, so Mistral Small 4 is 8.6 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Mistral Small 4 lands below the middle on ALE-Bench: 498, #59 of 79, with 73% of the field scoring higher. GPT-6 Astra (max) leads at 2,951; the closest result above is mimo-v2.5 at 514. It covers AtCoder heuristic-optimisation contests (performance rating). The board median is 764, so Mistral Small 4 is 266 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.

Strengths and weaknesses

Mistral Small 4 has too few results on boards with ten or more models to say where it is strong or weak. Treat any single score as a data point, not a verdict.

Price and value

The median input price across the 427 priced models is $0.500 per million tokens, so Mistral Small 4 costs less than the typical listing. Models that score within 1.5 points or better and cost less: Qwen2.5-Coder-7B-Instruct ($0.010, 49.5), DeepSeek V4.1 Flash (high) ($0.010, 48.9), Qwen2.5-Coder-3B-Instruct ($0.010, 46.5). Similar scores cost more on qwen3-235b-a22b ($0.180) and Llama 4 Maverick (FP8) ($0.188).

Release history

The Mistral AI model released before it was Devstral Small 2 on 2025-12-09, which scores 45.4. The next one was Mistral Medium 3.5 on 2026-04-28, at 43.5.

Reading these numbers

The 3 scores on this page come from 1 source: Epoch AI Benchmarking Hub (3). 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

#510 Qwen2.5-Coder-3B-Instruct (46.5)
#511 amazon.nova-pro-v1:0 (46.5)
#512 aya-expanse-32b (46.5)
#514 Qwen2-VL-7B-Instruct (46.5)
#515 Gemma 2 9B (46.5)
#516 deepseek-chat (46.5)