Back to Explore
Mistral: Mistral Medium 3.5
Mistral·Released 2026-04-30
Open Source262K ctxMoEMultimodal
About
Mistral Medium 3.5 is a dense 128B instruction-following model from Mistral AI. It supports text and image inputs with text output, and is designed for agentic workflows, coding, and complex...
Related Models
Pricing
Input
$1.50
per 1M tokens
Output
$7.50
per 1M tokens
Blended
$3.00
per 1M tokens
Cheaper than 20% of models. Median price is $0.54/1M tokens.
Cost Calculator
Tokens per day1M
100K100M
Daily
$3.00
Monthly
$90.00
vs. Similar Models
GPT-5.4 mini (medium)Q:-0.1
$1.69-44%
StepFun: Step 3.7 FlashQ:-0.2
$0.44-85%
Claude 4.5 Haiku (Reasoning)Q:-0.3
$2.00-33%
GPT-5.4 nano (medium)Q:+0.3
$0.46-85%
Performance
122
tokens/sec
Faster than 63% of models
0.57
seconds
Faster than 80% of models
17.01
seconds
Faster than 29% of models
Market Median
94 tok/s
29% faster
Median TTFT
1.11s
49% faster
Throughput/Dollar
41
tok/s per $/1M
Speed Comparison
inclusionAI: Ring-2.6-1T
123 tok/s+1%
Gemini 3 Pro Preview (low)
120 tok/s-1%
Qwen3 VL 30B A3B Instruct
120 tok/s-2%
Context Window
262K
tokens
Larger than 62% of models
Benchmarks
MMLU-ProNot evaluated
GPQA Diamond
74.8%
HLE
12.8%
LiveCodeBenchNot evaluated
SciCode
39.6%
TerminalBench Hard
33.3%
MATH-500Not evaluated
AIMENot evaluated
AIME 2025Not evaluated
IFBench
68.8%
Long Context Recall
61.0%
Tau2
94.2%
Market AverageTop Score