OpenGPTEvidence-based model decisionsStarting candidate 1
StableClaude Fable 5
claude-fable-5AnthropicUser preference · 1
01 · TASK RANKING
Everyday work
Drafting, analysis, planning, and mixed office tasks.
- Why these candidates
- Start with broadly capable models that have current public preference evidence, then test them on your own recurring work.
- What remains uncertain
- A general preference rank cannot predict accuracy on a specialized workflow.
Starting candidate
0 / 3Three official versions to test first
Public data narrows the field. It does not replace a controlled test on your own work. Select two or three candidates for a side-by-side public data comparison.
Starting candidate 2
StableGPT-5.6 Sol
gpt-5.6-solOpenAIUser preference · 11
Starting candidate 3
StableGemini 3.5 Flash
gemini-3.5-flashGoogleUser preference · 17
Public evidenceThe models use different public evidence scales. Missing evidence is not treated as zero, and OpenGPT does not combine unlike sources into one score.＋
Claude Fable 5
- Source rank
- 1
- Source value
- 1507
- Mapping reviewed
- Official identity mapping
- Official identity mapping ↗
- Research edition
- 2026-07 ·
GPT-5.6 Sol
- Source rank
- 11
- Source value
- 1485
- Mapping reviewed
- Official identity mapping
- Official identity mapping ↗
- Research edition
- 2026-07 ·
Gemini 3.5 Flash
- Source rank
- 17
- Source value
- 1476
- Mapping reviewed
- Official identity mapping
- Official identity mapping ↗
- Research edition
- 2026-07 ·
A practical starting point, not a universal winnerSelect two or three candidates for a side-by-side public data comparison.＋
The models use different public evidence scales. Missing evidence is not treated as zero, and OpenGPT does not combine unlike sources into one score.
A general preference rank cannot predict accuracy on a specialized workflow.
The evaluation opens with this task template and the first two selected models. You can edit every task before collecting answers.