NOUMENOpen app
NOUMEN

Qwen · Open weights

Qwen 3 235b A22B

Added Apr 29, 2025Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

Qwen 3 235b is a 235b model with 22B active parameters. Supports switching between thinking and non thinking: trigger thinking with /think and /no_think anywhere in a prompt or system message to toggle chain-of-thought reasoning.

  • Tools
  • Structured output
  • PDF input
Input
Text
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$0.0875
Output
$0.35
Cache read
$0.0175

Reference prices as published by the model provider. API example

Specifications

Context window
262.1K
Max output
16.4K
Average output (7 days)
616 tokens
Longer than 57% of 468 models

Side by side

Qwen 3 235b A22B compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Qwen 3 235b A22BQwenThis modelLFM2.5 2.6BliquidRemove
ProviderQwenliquid
Discovery tierOpen weightsOpen weights
AddedApr 29, 2025Aug 12, 2026
Limits
Context window262.14KBest128K
Max output16.38K32.77KBest
Average output616 tokens459 tokens
Pricing, USD per 1M tokens
Input$0.0875Best$0.10
Output$0.35$0.20Best
Cache read$0.0175—
Performance
Request success (24h)91.4%100.0%Best
Benchmark indexes
Intelligence Index8.38.4Best
Coding Index—7.7
Math Index23.7—
Capabilities
ReasoningNoYes
ToolsYesYes
Structured outputYesYes
PDF inputYesNo
Reasoning effort—minimal, low, medium, high, xhigh

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Qwen 3 235b A22B
  • LFM2.5 2.6B

Intelligence Index

  • Qwen 3 235b A22B8.3
  • LFM2.5 2.6B8.4

Qwen 3 235b A22B: Better than 13% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Qwen 3 235b A22B: —
    • LFM2.5 2.6B: 7.7

  • Math Index

    Composite of the math evaluations

    • Qwen 3 235b A22B: 23.7
    • LFM2.5 2.6B: —

Document reasoning

  • AA-LCRv1.1

    Reasoning across long documents

    • Qwen 3 235b A22B: 0.0%
    • LFM2.5 2.6B: 5.7%Top

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Qwen 3 235b A22B: 4.2%
    • LFM2.5 2.6B: 6.2%Top

  • GPQA Diamond

    Graduate-level science questions

    • Qwen 3 235b A22B: 61.3%Top
    • LFM2.5 2.6B: 55.8%

  • MMLU-Pro

    Multi-subject knowledge and reasoning

    • Qwen 3 235b A22B: 76.2%
    • LFM2.5 2.6B: —

  • IFBench

    Following precise output instructions

    • Qwen 3 235b A22B: 36.6%
    • LFM2.5 2.6B: —

Coding

  • SciCode

    Scientific computing in Python

    • Qwen 3 235b A22B: —
    • LFM2.5 2.6B: 14.4%

  • LiveCodeBench

    Recent competitive programming problems

    • Qwen 3 235b A22B: 34.3%
    • LFM2.5 2.6B: —

Math

  • AIME 2025

    Competition mathematics, 2025 problems

    • Qwen 3 235b A22B: 23.7%
    • LFM2.5 2.6B: —

  • AIME

    Competition mathematics

    • Qwen 3 235b A22B: 32.7%
    • LFM2.5 2.6B: —

  • MATH-500

    Competition math problems

    • Qwen 3 235b A22B: 90.2%
    • LFM2.5 2.6B: —

Legacy evaluations

  • Terminal-Bench Hard

    Earlier terminal task suite

    • Qwen 3 235b A22B: 6.1%
    • LFM2.5 2.6B: —

  • τ²-Bench Telecom

    Tool use in a customer-support simulation

    • Qwen 3 235b A22B: 27.2%
    • LFM2.5 2.6B: —

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Qwen 3 235b A22B: 0.0%
    • LFM2.5 2.6B: 5.7%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Prompt cache hits
0.0%
About $0.0875 per 1M input tokens with caching
Request success (24h)
91.3%
Unstable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 64.3%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-235b-a22b",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="qwen/qwen3-235b-a22b",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

Qwen

Other Qwen models, starting with the same family.