Alibaba Qwen · Benchmarks

Qwen3.6 27B benchmarks

Qwen3.6 27B has 10 published scores in this dataset, from a Alibaba Qwen release dated 2026-04. Its composite index is 85.8, which places it 12th of 22 ranked models and 7th among 15 open-weight models. Its strongest area is knowledge (index 99.3) and its weakest is coding (index 77.3). Its best placing is on AIME, where its 94.1 ranks 3rd of 5 models with a published score.

Deploy hereDeploy Qwen3.6 27B A4BOfficial page

You can self-deploy Qwen3.6 27B A4B on a dedicated GPU with OpenLLM Buddy and call it through an OpenAI-compatible API.

Category profile

Index 0–100: each score relative to the best published score on that test.

ReasoningCodingAgenticKnowledge
Reasoning
80.9#9 of 16 · 4 tests
Coding
77.3#10 of 13 · 4 tests
Agentic
—no data
Knowledge
99.3#2 of 4 · 2 tests

Model facts

Lab
Alibaba Qwen
Type
OpenLLM Buddy catalog
Released
2026-04
Parameters
27B dense
Context
262K (up to 1M)
License
Apache-2.0

Qwen3.6 27B hardware requirements and cost

At 4-bit it needs about 19 GB, which fits a single RTX 4090, so it can run on a home workstation with llama.cpp, Ollama, or vLLM. Full BF16 precision needs about 62 GB (1× H100 (80 GB)).

Estimated GPU memory for Qwen3.6 27B
PrecisionWeightsWith headroomSmallest setup
BF16 (full precision)54 GB62 GB1× H100 (80 GB)
FP8 / 8-bit27 GB31 GBRTX 5090 (32 GB)
4-bit (GGUF Q4_K_M)16 GB19 GBRTX 4090 (24 GB)

Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.

On OpenLLM Buddy: Qwen3.6 27B A4B, with an OpenAI-compatible API and no per-token fees. Prices start at $17 USD for 11 Hours.

  • 11 Hours: $17 USD
  • 24 Hours: $37 USD
  • 1 Week: $254 USD
  • 1 Month: $1013 USD
Deploy Qwen3.6 27B

Lighter or stronger open models

Open models ranked just above and below on the composite index, with their 4-bit memory needs.

Qwen3.6 27B speed and pricing

Measured figures link to their source; estimates are labelled.

API output speed
57 tok/sArtificial Analysis median · Artificial Analysis
Time to first token
0.65 sOpenRouter model page provider stats
API price per 1M tokens
$0.450 in · $2.70 outlab's own endpoint on OpenRouter · OpenRouter endpoints API
On your own RTX 4090 (4-bit)estimate
28–39 tok/s

Estimate for one user on one GPU: active weights read per token ÷ memory bandwidth at 50–70% efficiency, plus 3 ms overhead per token. Real speed varies with engine, context length and quantization; serving several users at once raises total throughput.

Compare speed and price across all models

Score by test

Rank is among every model with a published score on that test.

GPQA Diamond87.8%Graduate-level, Google-proof science questions in biology, physics, and chemistry.#15 of 19 · 8.3 behind GPT-6 AstraHumanity's Last Exam24.0%Expert-written closed-ended questions at the frontier of human knowledge.#20 of 23 · 40.4 behind Claude Opus 5.5AIME94.1%AIME 2026#3 of 5 · 3.6 behind Reflection Beam 1.0HMMT February93.8%HMMT Feb 2025#2 of 2 · 1.4 behind DeepSeek V4 ProLiveCodeBench83.9%LiveCodeBench v6#5 of 7 · 9.6 behind DeepSeek V4 ProSWE-bench Verified77.2%Real GitHub issues from popular Python repositories, fixed end to end.#3 of 5 · 4.4 behind Reflection Beam 1.0SWE-bench Pro53.5%Harder, longer software tasks across many languages and repositories.#9 of 9 · 36.4 behind Claude Opus 5.5Terminal-Bench59.3%Terminal-Bench 2.0#12 of 13 · 31.3 behind DeepSeek V4.1 FlashMMLU-Pro86.2%Broad academic knowledge with ten answer options instead of four.#3 of 5 · 1.3 behind DeepSeek V4 ProMMMU82.9%MMMU#1 of 1 · top score

Nearby in the composite ranking

Models with the closest composite index.

  1. 1Muse Spark Open88.0
  2. 2Reflection Beam 1.087.0
  3. 3GPT-6.1 Sol86.5
  4. 4Qwen3.6 27B85.8
  5. 5Qwen3.8 27B83.2
  6. 6GLM 5.382.7
  7. 7DeepSeek V4.1 Flash79.8

Head-to-head comparisons

Side-by-side pages built from the tests both models report.

Sources

ModelTestScoreSettingSource
Qwen3.6 27BGPQA Diamond87.8—Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BHumanity's Last Exam24.0—Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BAIME94.1AIME 2026Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BHMMT February93.8HMMT Feb 2025Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BLiveCodeBench83.9LiveCodeBench v6Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BSWE-bench Verified77.2—Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BSWE-bench Pro53.5—Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BTerminal-Bench59.3Terminal-Bench 2.0Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BMMLU-Pro86.2—Qwen3.6-27B model card (Hugging Face)
Qwen3.6 27BMMMU82.9MMMUQwen3.6-27B model card (Hugging Face)

Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works