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

Qwen3.7-Max×

DeepSeek-V4-Flash×

Qwen3.6-27B×

Llama 4 Scout×Clear all

	Qwen3.7-Max Alibaba	DeepSeek-V4-Flash DeepSeek	Qwen3.6-27B Alibaba	Llama 4 Scout Meta
Vendor	Alibaba	DeepSeek	Alibaba	Meta
Family	Qwen	DeepSeek	Qwen	Llama
Release date	2026-05-20	2026-04-22	2026-04-22	2025-04-05
Context window	1,000,000 tokens	1,000,000 tokens	262,144 tokens	10,000,000 tokens
Parameters	—	284B (13B active)	27B (dense)	109B
Modality	text	text	text, vision, video	text, vision
License	proprietary	MIT	Apache-2.0	Llama 4 Community License
Source	proprietary	open weights	open weights	open weights
Description	Alibaba's flagship agent model — 1M-token context, extended-thinking mode, 56.6 on the Artificial Analysis Intelligence Index v4.0 (5th overall, #1 Chinese). 50.8% on Terminal-Bench Hard. Designed for long-horizon agent workloads (hundreds-to-thousands of steps). Closed-weight, $2.50/$7.50 per 1M tokens.	Smaller, faster sibling to DeepSeek-V4-Pro. Same 1M context window with a much lighter 284B / 13B-active MoE.	Alibaba's first dense open-weight in the 3.6 family. Strong agentic-coding scores (77.2 SWE-bench Verified, matching Claude 4.5 Opus on Terminal-Bench 2.0). Supports 201 languages and multimodal text/image/video input.	—
Links	vendor →	weights →	weights →	weights →
Benchmarks
MMLU-Pro	83.7%	—	78.9%	—
GPQA-D	83.0%	—	—	—
HumanEval	93.9%	—	91.5%	—
Aider	78.4%	—	70.2%	—
AIME-25	90.4%	—	—	—
LiveCB	71.0%	—	—	—
MMMU	—	—	70.3%	—