Compare models

Pick up to 4models. Specs render side-by-side. Share the URL — it's stateless.

Comparing agents instead? Switch to agent compare →

Selected

DeepSeek-V4-Pro×

Gemini 3.5×

Mistral Large 2×

	DeepSeek-V4-Pro DeepSeek	Gemini 3.5 Google	Mistral Large 2 Mistral	o3 OpenAI
Vendor	DeepSeek	Google	Mistral	OpenAI
Family	DeepSeek	Gemini	Mistral	o-series
Release date	2026-04-22	2026-05-20	2024-07-24	2025-04-16
Context window	1,000,000 tokens	2,000,000 tokens	128,000 tokens	200,000 tokens
Parameters	1.6T (49B active)	—	123B	—
Modality	text	text, vision, audio, video	text	text, vision
License	MIT	proprietary	Mistral Research License	proprietary
Source	open weights	proprietary	open weights	proprietary
Description	DeepSeek's flagship open-weight MoE. 1.6T parameters with 49B activated, 1M-token context, and a hybrid attention scheme (CSA + HCA) that delivers long-context inference at ~27% of V3.2's FLOPs.	Google DeepMind's next-gen Gemini — positioned as "frontier intelligence with action". Built for complex agentic workflows. Announced at Google I/O 2026.	Large text-only Mistral model with a 128K context window and 123B parameters, tuned for strong instruction following and long-context reasoning. Mistral's flagship open-weights release in the Large 2 line.	Reasoning-focused model in the o-series.
Links	weights →	vendor →	vendor →weights →
Benchmarks
MMLU-Pro	84.2%	88.9%	—	—
GPQA-D	82.4%	91.2%	—	87.7%
HumanEval	95.1%	—	89.0%	—
Aider	80.1%	—	—	—
AIME-25	88.6%	96.1%	—	88.9%
LiveCB	72.4%	—	—	—
MMMU	—	86.0%	—	—