Qwen: Qwen3 30B A3B Thinking 2507

Qwen

qwen/qwen3-30b-a3b-thinking-2507

Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking. The model is designed specifically for “thinking mode,” where internal reasoning traces are separated...

Context

82K

Input /1M

$0.20

Output /1M

$2.40

Latency (TTFT)

333ms

Pricing

Per 1M tokens, in USD

Input$0.20

Output$2.40

Specs

text intext out

TokenizerQwen3

Max output tokens32,768

Supported parameters

frequency_penaltyinclude_reasoningmax_tokenspresence_penaltyreasoningresponse_formatseedstoptemperaturetool_choicetoolstop_ktop_p

Where it ranks on price

Against 157 routable models

Input$0.20

Cheaper than 76% of routable models

Output$2.40

Cheaper than 49% of routable models

Benchmarks

Intelligence9.8

Coding12.1

Agentic0.9

Composite scores (0–100) from Artificial Analysis, via OpenRouter's model catalog.

Call this model

Swap in your Mizan key — the request/response bodies are OpenAI chat-completion shaped.

cURL

curl https://api.app-mizan.com/v1/chat/completions \
  -H "Authorization: Bearer mizan_rt_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-30b-a3b-thinking-2507",
    "messages": [{ "role": "user", "content": "Say hi in five words." }]
  }'

Models

Qwen: Qwen3 30B A3B Thinking 2507

Qwen

qwen/qwen3-30b-a3b-thinking-2507

Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking. The model is designed specifically for “thinking mode,” where internal reasoning traces are separated...

Context

82K

Input /1M

$0.20

Output /1M

$2.40

Latency (TTFT)

333ms

Pricing

Per 1M tokens, in USD

Input$0.20

Output$2.40

Specs

text intext out

TokenizerQwen3

Max output tokens32,768

Supported parameters

frequency_penaltyinclude_reasoningmax_tokenspresence_penaltyreasoningresponse_formatseedstoptemperaturetool_choicetoolstop_ktop_p

Where it ranks on price

Against 157 routable models

Input$0.20

Cheaper than 76% of routable models

Output$2.40

Cheaper than 49% of routable models

Benchmarks

Intelligence9.8

Coding12.1

Agentic0.9

Composite scores (0–100) from Artificial Analysis, via OpenRouter's model catalog.

Call this model

Swap in your Mizan key — the request/response bodies are OpenAI chat-completion shaped.

cURL

curl https://api.app-mizan.com/v1/chat/completions \
  -H "Authorization: Bearer mizan_rt_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-30b-a3b-thinking-2507",
    "messages": [{ "role": "user", "content": "Say hi in five words." }]
  }'