Qwen3.6 27B

Dense 27B that runs on one 24GB GPU.

Released

Apr 22, 2026

Price in / out

—

Blended $/1M

—

Context

131K

Parameters: 27B. License: Apache-2.0.

Benchmark scoresvs Qwen3.6 Plus

Benchmark	Score	Δ vs prior	Rank (current)	Configuration	Source
SWE-bench Verified	77.2%	-1.6	#5 of 9	leaderboard	benchlm.ai ↗
Humanity's Last Exam	24%	-4.8	#15 of 15	leaderboard	benchlm.ai ↗
AIME 2026	94.1%	-1.2	#2 of 2	leaderboard	benchlm.ai ↗
τ²-bench (Telecom)	94.2%	-3.5	#3 of 10	leaderboard	benchlm.ai ↗
MMLU-Pro	86.2%	-2.3	#3 of 4	leaderboard	benchlm.ai ↗
Artificial Analysis Intelligence Index	37	-2.6	#7 of 11	leaderboard	benchlm.ai ↗

Qwen versions

Timeline

Version	Released	$ in / out	Terminal-Bench	SWE-bench	GPQA	HLE	AIME	τ²-bench
Qwen3.5 397B-A17BSuperseded	2026-02-15	—	—	76.2%	—	28.7%	93.3%	83.9%
Qwen3.6 PlusSuperseded	2026-04-01	—	53.2%	78.8%	—	28.8%	95.3%	97.7%
Qwen3.6 27BSuperseded	2026-04-22	—	—	77.2%	—	24%	94.1%	94.2%
Qwen3.7 MaxSuperseded	2026-05-20	$2.5 / $7.5	61%	80.4%	92.4%	41.4%	—	94.7%
Qwen3.7 PlusSuperseded	2026-06-01	—	52.8%	77.7%	90.3%	34.7%	—	93%
Qwen3.8 Max	2026-08-02	$2 / $6	67.4%	—	92.6%	43.6%	—	—
Qwen3.8 27B	2026-08-05	—	58.4%	—	89.2%	30.8%	—	—
Qwen3.8 Flash-Next	2026-08-24	—	—	—	91.7%	35.9%	—	—

Compare

Open the Comparison Lab with Qwen3.6 27B and its open-weight peers pre-selected, then toggle any benchmark or model.

Every Qwen version over time

Shipping with it?

Benchmarks are a starting point. If you run Qwen3.6 27B in production, post what it does well and where it falls over; other founders are deciding right now.

Write about Qwen3.6 27B

Current open-weight peers

All

gpt-oss-120bOpenAI Gemma 4 31BGoogle DeepMind Muse Glimmer 30BMeta DeepSeek V4.1 Flash$0.14 in DeepSeek V4 Pro$0.435 in Kimi K3$3 in