New: Qwen4-Max eval report is here.Read the report ↗
Updated: Data as of:

Claude Fable 5.2 vs Qwen4-Max

Find the right model, side by side.

Across 7 shared evaluation tracks, Claude Fable 5.2 leads on 7, Qwen4-Max leads on 0, and 0 are tied. These counts describe reported scores within matched setups, not an overall model ranking. At the initially selected providers, input/output prices are $8.00 / $40.00 for Claude Fable 5.2 and $1.50 / $4.50 for Qwen4-Max, per million tokens.

First model
Second model

Model specifications

Specifications for Claude Fable 5.2 and Qwen4-Max. Prices and performance use the selected provider offering.
SpecificationsClaude Fable 5.2AnthropicQwen4-MaxQwen
Provider
Context sizetokens1,000,0001,000,000
Cutoff date2026-06Not reported
Input costUSD / 1M tokens$8.00$1.50
Output costUSD / 1M tokens$40.00$4.50
Max outputtokens128,000262,144
LatencyTTFT · secondsNot reportedNot reported
Speedtokens / sec103.417Not reported

Blended token cost

USD per 1M tokens · 50% input + 50% output · lower is better

Blended token cost. Claude Fable 5.2: $24.00; Qwen4-Max: $3.00$24.00Claude Fable 5.2$3.00Qwen4-Max

Output speed by effort

End-to-end output tokens per second · higher is better

Scroll to see all efforts →
Output speed by effort. Claude Fable 5.2 (low): 92.4; Claude Fable 5.2 (medium): 88.4; Claude Fable 5.2 (high): 93; Claude Fable 5.2 (xhigh): 103.4; Claude Fable 5.2 (max): 102.8; Qwen4-Max: Not reported92.4Low88.4Medium93High103.4Xhigh102.8MaxNot reportedUnreported

First-party APIs. Five-day workload averages, checked November 10, 2026. Methodology ↗

  • Qwen4-Max: No effort-level measurements available.

Latency by effort

Seconds until the first answer, including thinking · lower is better

Scroll to see all efforts →
Latency by effort. Claude Fable 5.2 (low): 8.73; Claude Fable 5.2 (medium): 13.45; Claude Fable 5.2 (high): 30.79; Claude Fable 5.2 (xhigh): 73.44; Claude Fable 5.2 (max): 235.06; Qwen4-Max: Not reported8.73Low13.45Medium30.79High73.44Xhigh235.06MaxNot reportedUnreported

First-party APIs. Five-day workload averages, checked November 10, 2026. Methodology ↗

  • Qwen4-Max: No effort-level measurements available.

Benchmark performance

7 shared benchmark comparisons. Versions and setups matched.

Claude Fable 5.2Qwen4-Max

Agent Index

Composite of 7 benchmarks · score out of 100 · higher is better

View index ↗
Agent Index. Claude Fable 5.2: 89.6; Qwen4-Max: Not reported89.6Claude Fable 5.2Not reportedQwen4-Max
  • Qwen4-Max: Needs all 7 benchmark results (4/7 available).