AI Index

Qwen3-0.6B

Language Alibaba Qwen Partly verified Released Apr 29, 2025
  • Reasoning
  • Reasoning can be off
  • Tool use
  • Structured output
  • Open weights

Prices

1 offering · prices as listed
Alibaba Cloud Model Studio Official Global · USD qwen3-0.6b
Conditions/ 1M tokensInputOutput
Standard no longer listed since Oct 1, 2026 past price, from the vendor’s page archived Jan 16, 2026$0.11$0.42

Prices from Alibaba Cloud Model Studio pricing (checked Sep 30, 2026)

Standard: the price a request with no conditions pays (the headline price elsewhere on the site).

Details

Licence
apache-2.0
Family
qwen
Model ID

Not stated on Alibaba Qwen’s pages: max input, input modalities, output modalities, reasoning efforts, max thinking budget, built-in tools, knowledge cutoff, aliases.

Run locally

Weights: Qwen/Qwen3-0.6B
  • Parameters 0.8B
  • Dense
  • Native precision BF16
  • Layers 28
  • KV cache per token 112 KB
  • Config positions 40K (config.json max_position_embeddings)
  • API context 250K (maker’s API)
My hardware GeForce RTX 4090 24 GB · 1008 GB/s · 64 GB DDR5 dual channel · 89.6 GB/s
On your hardware · GeForce RTX 4090
Runs well BF16 (native) (automatic) 8K context 206–309 tok/sOne GPU
WeightsKV cacheOverheadUsable
Needs
≈ 2.8 GB (weights 1.4 (maker’s file) + KV cache 0.9 + overhead 0.5)
Usable
21.6 GB (90% of its memory)
Generation speed
206–309 tok/s
Longest context
40K (the model’s limit)
First token
about 0.02–0.04 s for a 2K prompt
PrecisionWeightsNeeds at 8KOn your hardwareLongest context
BF16 (native)maker’s file1.4 GB≈ 2.8Runs well 206–309 tok/s40K
Q8_0maker’s file0.6 GB≈ 2.0Runs well 319–479 tok/s40K
INT8 (GPTQ)maker’s file0.7 GB≈ 2.1Runs well 296–444 tok/s40K
FP8maker’s file1.0 GB≈ 2.4Runs well 252–378 tok/s40K
Q6_K0.6 GB≈ 1.9Runs well 324–486 tok/s40K
Q5_K_M0.5 GB≈ 1.9Runs well 342–513 tok/s40K
Q5_00.5 GB≈ 1.9Runs well 346–519 tok/s40K
Q4_K_M0.4 GB≈ 1.8Runs well 361–541 tok/s40K
Q4_00.4 GB≈ 1.8Runs well 370–555 tok/s40K
IQ4_XS0.4 GB≈ 1.7Runs well 376–565 tok/s40K
Q3_K_Mquality drops0.3 GB≈ 1.7Runs well 386–578 tok/s40K
IQ3_XXSquality drops0.3 GB≈ 1.6Runs well 411–616 tok/s40K
Q2_Kquality drops0.3 GB≈ 1.6Runs well 413–619 tok/s40K
IQ2_XXSquality drops0.2 GB≈ 1.6Runs well 445–667 tok/s40K

Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.

Compared with other hardware

Each at the highest precision it runs. Click one for details.

Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 3080; unified memory Apple M5 · 16 GB.

Run commands

vLLM
vllm serve Qwen/Qwen3-0.6B --max-model-len 8192
Memory to fine-tune
Full≈ 13.0 GBFits your hardware
LoRA≈ 3.3 GBFits your hardware
QLoRA≈ 2.3 GBFits your hardware

Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.

Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated

Cited scores

From third-party leaderboards, not measured by us

This model isn’t on any leaderboard we cite yet.

Changes

All changes →

No changes recorded yet.

Sources

Alibaba Qwen’s official policies: data use, privacy, terms →

Some values on this page are still waiting to be checked against official pages: a filled dot is verified, a hollow one is not yet (with or without a source on file). How we verify data

Alibaba Qwen · China · qwen.ai

Snapshots, the same family and other models from this provider