models.fyiPreview data

Qwen3-235B-A22B

AlibabaOpen weights

Alibaba's open MoE workhorse — mid-frontier quality, extremely cheap on third-party hosts.

Index @ Medium

66.7

Context

128K

$/M in · out

$0.70 · $2.80

Median speed

70 tok/s

Released

Jul 21, 2025

02·Which effort for which task

The sweet spot, per task

Cheapest effort tier that keeps ≥96% of this model's peak quality on the benchmarks behind each task type. Below the floor number, quality is not okay — it just fails quietly.

Everyday writing & Q&A

Medium

Chat, drafting, summaries, general questions.

keeps 98.0% of peak−66% cost

Code assist

High

Autocomplete, single functions, small fixes.

keeps 100.0% of peakpeak tier needed

Coding agents

High

Real repos, multi-file changes, end-to-end tickets.

keeps 100.0% of peakpeak tier needed

Math & quantitative

High

Derivations, statistics, financial models.

keeps 100.0% of peakpeak tier needed

⚠ at Low: only 70% of peak — avoid

Hard research problems

High

Novel, frontier, genuinely difficult questions.

keeps 100.0% of peakpeak tier needed

⚠ at Low: only 78% of peak — avoid

Agents & tool use

High

Long workflows with tools, policies, and state.

keeps 100.0% of peakpeak tier needed

03·The effort ladder

Quality and cost, tier by tier

On everyday tasks, Medium keeps 98.0% of High quality at $0.0080 vs $0.023 per task.

Intelligence Index by effort

Cost per typical task by effort

04·Benchmark detail

Scores at Medium

Low Medium High

at Mediumat High (peak)

Benchmark	Medium	Peak	What it predicts
MMLU-Pro	80.9	83.0	Everyday Q&A · Drafting emails & docs · Summarizing articles
GPQA Diamond	74.3	81.0	Technical analysis · Root-cause debugging · Scientific writing
Humanity's Last Exam	16.5	18.0	Novel research questions · Frontier problem solving
AIME 2025	72.1	86.0	Financial modeling · Statistics · Algorithm design
SWE-bench Verified	63.1	68.0	Coding agents · Bug fixing in real repos · Multi-file refactors
LiveCodeBench	70.5	76.0	Writing functions from scratch · Algorithms · Code autocomplete
τ²-bench	67.8	74.0	Tool-using agents · Customer-facing automation · Multi-step workflows
IFEval	85.7	87.0	Structured output (JSON) · Templated generation · Batch pipelines

05·Host variants

Same weights, different model

Hosts serve Qwen3-235B-A22B at different quantizations and stacks. Index below is measured per host at Medium effort — the headline number only applies to the reference serving.

Host	Quant	Index	Δ vs reference	$/M in · out	tok/s	TTFT	Uptime	Gateway id
Alibaba (official)reference	bf16	66.7	—	$0.70 · $2.80	70	0.80s	99.1%	qwen3-235b
Fireworks	fp8	66.2	-0.5	$0.22 · $0.88	190	0.35s	99.0%	qwen3-235b@fireworks
DeepInfra	fp8	65.7	-1.0	$0.13 · $0.60	130	0.50s	98.3%	qwen3-235b@deepinfra

06·Drift

Is it still the same model this week?

Alibaba (official)FireworksDeepInfra

Weekly Intelligence Index · dashed rules mark serving-change events

No serving-change events detected for Qwen3-235B-A22B in the tracked window.

07·Call it

One key, this model, your effort

The gateway speaks the OpenAI Chat Completions dialect. Pick a host with qwen3-235b@host syntax, set reasoning_effort, and the sweet spot above becomes one line of config. Full docs →

chat with Qwen3-235B-A22B

curl https://models.fyi/api/v1/chat/completions \
  -H "Authorization: Bearer $MFYI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-235b",
    "reasoning_effort": "medium",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
# local dev: replace https://models.fyi with http://localhost:3000