WANDR
Perplexity agent evaluation on large-scale data collection tasks.
- Category
- Tool use
- Unit
- Percent (%)
- Higher is better
- Comparison groups
- 1total
- Currently available model rankings · Ranked results: 5
- All models (records): Groups: 1 · Entries: 15
Rankings
Values with different source types (official, third-party, paper) or measurement conditions cannot be compared, so rankings are separated by comparison group (the same source, or sources verified to use the same conditions).
Rankings include only models currently available in Claude Code / Codex, excluding legacy models.
- View on rankings page (by metric)
- Cost balance (frontier and effort changes)
- Overall ranking (performance first)
official results (Introducing Claude Opus 5.5)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Max | 72.3% | $37.92Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Max: 72.3% $37.92" |
| 2 | Claude Opus 5.5 | Xhigh | 71.3% | $29.06Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Xhigh: 71.3% $29.06" |
| 3 | Claude Opus 5.5 | High | 67.3% | $16.88Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · High: 67.3% $16.88" |
| 4 | Claude Opus 5.5 | Med | 62.8% | $11.20Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Med: 62.8% $11.20" |
| 5 | Claude Opus 5.5 | Low | 31.2% | $1.20Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Low: 31.2% $1.20" |
All model rankings (records)
These rankings are recalculated by comparison group including legacy models and models unavailable in Claude Code / Codex. They differ from the rankings above because the set of models differs.
- View all model rankings (by metric)
- Cost balance (all models, records)
- Overall ranking (all models, records)
Show rankings including legacy and unavailable models (entries: 15)
official results (Introducing Claude Opus 5.5)
| Rank | Model | effort | Score | Source cost (metric) | measured/reported by | source |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 5.5 | Max | 72.3% | $37.92Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Max: 72.3% $37.92" |
| 2 | Claude Opus 5.5 | Xhigh | 71.3% | $29.06Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Xhigh: 71.3% $29.06" |
| 3 | Claude Fable 5.1 | Max | 68.7% | $49.02Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Fable 5.1 · Max: 68.7% $49.02" |
| 4 | Claude Fable 5.1 | Xhigh | 67.7% | $42.81Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Fable 5.1 · Xhigh: 67.7% $42.81" |
| 5 | Claude Opus 5.5 | High | 67.3% | $16.88Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · High: 67.3% $16.88" |
| 6 | Claude Opus 5 | Max | 67.2% | $61.62Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5 · Max: 67.2% $61.62" |
| 7 | Claude Opus 5 | Xhigh | 67.0% | $53.84Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5 · Xhigh: 67.0% $53.84" |
| 8 | Claude Fable 5.1 | High | 66.7% | $33.62Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Fable 5.1 · High: 66.7% $33.62" |
| 9 | Claude Opus 5 | High | 64.6% | $43.49Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5 · High: 64.6% $43.49" |
| 10 | Claude Fable 5.1 | Med | 64.5% | $27.41Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Fable 5.1 · Med: 64.5% $27.41" |
| 11 | Claude Fable 5.1 | Low | 63.3% | $23.25Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Fable 5.1 · Low: 63.3% $23.25" |
| 12 | Claude Opus 5.5 | Med | 62.8% | $11.20Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Med: 62.8% $11.20" |
| 13 | Claude Opus 5 | Med | 58.1% | $23.97Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5 · Med: 58.1% $23.97" |
| 14 | Claude Opus 5 | Low | 50.5% | $10.65Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5 · Low: 50.5% $10.65" |
| 15 | Claude Opus 5.5 | Low | 31.2% | $1.20Cost per attempt | Anthropic | Introducing Claude Opus 5.5Retrieved 2026-11-12 · Vendor-reported value. A point in the figure "WANDR / Accuracy vs Cost" on the Anthropic announcement page. values were taken from each point's aria-label (DOM); the page shows them only in the figure; value for its own model (announced by Anthropic); WANDR: Claude was run with offline web search/fetch, programmatic tool calling, code execution, and a 980k token task budget (different from the Perplexity public configuration)Source excerptWANDR — Accuracy vs Cost: "Opus 5.5 · Low: 31.2% $1.20" |