AI Model Release Tracker

Qwen3Open Weight

Released

Qwen3 is an AI model released by Qwen on Tuesday, Apr 29 2025, 54 days after QwQ-32B. It is an open-weight model — the trained weights are available to download and run. It was released in 0.6B and 1.7B and 4B and 8B and 14B and 30B and 32B and 235B parameter sizesThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality.. Benchmark results (shown below) cover BullshitBench v2.

Available from

NextBit14b—$0.10$0.2241Kint4
DeepInfra14b—$0.12$0.2441Kfp8
DeepInfra32b—$0.08$0.2841Kfp8
DekaLLM30b-a3b-instruct-2507—$0.09$0.30256K—
SiliconFlow30b-a3b-instruct-2507—$0.09$0.30256Kfp8
GMICloud235b-a22b-2507—$0.0875$0.35256Kfp8
DeepInfra30b-a3b—$0.12$0.5041Kfp8
DeepInfra235b-a22b-2507—$0.09$0.55256Kfp8
SiliconFlow32b—$0.14$0.57128Kfp8
Nebius235b-a22b-2507—$0.20$0.60256Kfp8
Venice235b-a22b-2507—$0.15$0.75128Kfp8
Parasail235b-a22b-2507—$0.14$0.80128Kfp8
Google235b-a22b-2507us-south1$0.25$1.00256K—
Venice235b-a22b-thinking-2507—$0.45$3.50128Kfp8
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked November 3, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

BullshitBench v2Nonsense detection — Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better.
6%
#93 of 98

Compare Qwen3 with

Qwen3

Suggested comparisons

Qwen3vsQwen3.8-Max-0902Qwen3vsGPT-6.1 TerraQwen3vsClaude Haiku 5.5Qwen3vsGemini 4 CarbonQwen3vsMuse Spark OpenQwen3vsGrok 4.7Qwen3vsDeepSeek-V4.1-Flash-1017Qwen3vsMistral Large 4Qwen3vsKimi K3.1 ProQwen3vsGLM-5.3-FlashQwen3vsNemotron 3.5 Bolt E4B 4GB

About

Qwen3 was introduced on April 29, 2025 with eight initial open-weight models under Apache 2.0. The lineup included six dense sizes from 0.6B to 32B and two mixture-of-experts models, 30B-A3B and 235B-A22B. The initial checkpoints combined thinking and non-thinking behavior in the same weights.

The release emphasized multilingual coverage and reasoning for mathematics, code and tool use. Developers could switch thinking modes and manage the computation spent on a response. The initial hybrid models should be distinguished from later Qwen3 checkpoints released specifically as Instruct or Thinking variants.

Frequently asked questions

Qwen3 was released by Qwen on Tuesday, Apr 29 2025.

All Qwen releases

24 tracked