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

Mistral Large 2×

Gemini 3 Pro×

DeepSeek-V4-Pro×

Llama 4 Maverick×Clear all

	Mistral Large 2 Mistral	Gemini 3 Pro Google	DeepSeek-V4-Pro DeepSeek	Llama 4 Maverick Meta
Vendor	Mistral	Google	DeepSeek	Meta
Family	Mistral	Gemini	DeepSeek	Llama
Release date	2024-07-24	2026-04-22	2026-04-22	2025-04-05
Context window	128,000 tokens	2,000,000 tokens	1,000,000 tokens	1,000,000 tokens
Parameters	123B	—	1.6T (49B active)	400B
Modality	text	text, vision, audio, video	text	text, vision
License	Mistral Research License	proprietary	MIT	Llama 4 Community License
Source	open weights	proprietary	open weights	open weights
Description	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.	Google DeepMind's flagship multimodal model. Best-in-class for multimodal understanding and agentic / vibe coding workflows. A "Deep Think" reasoning mode is available to AI Ultra subscribers.	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.	Mixture-of-experts open-weight model from Meta.
Links	vendor →weights →	vendor →	weights →	weights →
Benchmarks
MMLU-Pro	—	87.1%	84.2%	80.5%
GPQA-D	—	88.6%	82.4%	—
HumanEval	89.0%	96.5%	95.1%	—
Aider	—	84.8%	80.1%	—
AIME-25	—	92.8%	88.6%	—
LiveCB	—	76.1%	72.4%	—
MMMU	—	84.3%	—	68.5%