← Back to model catalog

VENICE ATLAS · MODEL DOSSIER · MISTRAL SMALL 4

Model overview

Mistral Small 4

Mistral Small 4 unifies instruction following, reasoning, coding, and vision in a single 119B MoE model with 256K context and configurable reasoning effort.

mistral-small-2603 · venice.ai
Text · Venice catalog$0.19 per 1M tokensPrice on Venice AI
ModalityTextOn Venice AI
Primary price$0.19 per 1M tokensper 1M tokens
Context window256KMax tokens
Benchmarks9Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

General intelligence1 result
Arena Hardarena-hard · general
58.3%
ReportedSep 2026
Instruction Following1 result
IFBenchifbench · instruction_following
48.0%
ReportedSep 2026
Language2 results
COLLIEcollie · language
62.9%
ReportedSep 2026
MMLU-Prommlu-pro · language
78.0%
ReportedSep 2026
Math1 result
AIME 2025aime-2025 · math
83.8%
ReportedSep 2026
Reasoning3 results
AA-LCRaa-lcr · reasoning
71.2%
ReportedSep 2026
GPQAgpqa · reasoning
71.2%
ReportedSep 2026
LiveCodeBenchlivecodebench · reasoning
63.6%
ReportedSep 2026
Vision1 result
MMMU-Prommmu-pro · vision
60.0%
ReportedSep 2026

Learn what each benchmark measures ↗

Related models

More from mistral