{"schema":"frankx/arena-receipts-manifest@1","_description":"Index of first-party model-arena eval receipts published by frankx.ai. Every measurement claim on /research/model-arena traces to one of these files. Figures sourced from third parties live in the external snapshot and carry their own licence; do not merge the two into a ranking.","lastMeasured":"2026-08-28","receiptCount":7,"plannedEvaluations":[{"id":"image-workflows-v1","status":"not_run","protocol":"/research/image-workflows/protocol.json","guide":"/guides/reference-to-campaign","measuredRuns":0}],"receipts":[{"roundId":"arena-2026-08-28-r5-deep-reasoning","date":"2026-08-28","title":"round-5-deep-reasoning: 6 tasks across 5 families where one wrong intermediate step propagates; fully mechanical verification; per-run token and cost accounting","contestants":["claude-fable-5","claude-opus-5","claude-sonnet-5","claude-haiku-4-5"],"taskCount":6,"href":"/research/arena-receipts/2026-08-28-r5-deep-reasoning.json"},{"roundId":"arena-2026-06-12-grok-composer25-model-lane","date":"2026-06-12","title":"grok-composer-special: 2 harness-relevant agentic tasks (sustained instruction following + multi-step verification + repo-grounded output discipline). Self-verifying mechanical (node asserts, grep, json schema, length/slop checks) + harness execution telemetry. Visual composer test via image_gen (routing card). External web evals grounding for cross-family context.","contestants":["grok-4.3"],"taskCount":2,"href":"/research/arena-receipts/2026-06-12-grok-composer25-model-lane.json"},{"roundId":"arena-2026-06-10-r3-lineup-4way","date":"2026-06-10","title":"round-3-lineup: full Anthropic lineup (Fable 5 / Opus 4.8 / Sonnet 4.6 / Haiku 4.5), 4 tasks, fully mechanical verification — zero LLM-judge dependence","contestants":["claude-fable-5","claude-opus-4-8","claude-sonnet-4-6","claude-haiku-4-5-20251001"],"taskCount":4,"href":"/research/arena-receipts/2026-06-10-r3-lineup-4way.json"},{"roundId":"arena-2026-06-10-r4-work-samples","date":"2026-06-10","title":"Round 4: Work Samples (2026-06-10)","contestants":["claude-fable-5","claude-opus-4-8"],"taskCount":2,"href":"/research/arena-receipts/2026-06-10-r4-work-samples.json"},{"roundId":"arena-2026-06-09-fable5-vs-opus48","date":"2026-06-09","title":"Round 1: Fable 5 vs Opus 4.8 (2026-06-09)","contestants":["claude-fable-5","claude-opus-4-8"],"taskCount":4,"href":"/research/arena-receipts/2026-06-09-fable5-vs-opus48.json"},{"roundId":"arena-2026-06-09-r2-stress","date":"2026-06-09","title":"round-2-stress: behavioral traps specific to the SIS multi-agent setup; all tasks mechanically checkable, zero LLM-judge dependence","contestants":["claude-fable-5","claude-opus-4-8"],"taskCount":5,"href":"/research/arena-receipts/2026-06-09-r2-stress-fable5-vs-opus48.json"},{"roundId":"arena-2026-06-09-r3-true-challenge","date":"2026-06-09","title":"Round 3: True Challenge (2026-06-09)","contestants":["claude-fable-5","claude-opus-4-8"],"taskCount":4,"href":"/research/arena-receipts/2026-06-09-r3-true-challenge.json"}],"externalSnapshot":{"schema":"frankx/external-intelligence@1","generatedAt":"2026-10-01T13:18:19.210Z","seeded":false,"sources":[{"id":"models-dev","status":"ok","licenseState":"confirmed","license":"MIT","retrievedAt":"2026-10-01T13:18:19.210Z"},{"id":"openrouter-models","status":"unverified","licenseState":"unverified","license":"OpenRouter ToS — redistribution not yet verified","retrievedAt":null},{"id":"openrouter-rankings","status":"unverified","licenseState":"unverified","license":"OpenRouter rankings terms — redistribution not yet verified","retrievedAt":null}],"linkOnly":[{"id":"artificial-analysis","url":"https://artificialanalysis.ai/","reason":"Free-tier terms are treated as no-redistribution pending a commercial licence."},{"id":"swe-bench","url":"https://www.swebench.com/","reason":"CC BY-NC 4.0 is not compatible with redistribution on this commercial site."},{"id":"lmarena","url":"https://lmarena.ai/leaderboard","reason":"Ingest remains deferred until the dataset licence and Parquet path are verified end to end."}]},"invalidReceipts":[]}