HiveBase routes each work class to a preferred model — and for coding, to a preferred harness × model × role. This page publishes that opinion, with the evidence we are allowed to show.
Tracked sources are named and linked. Artificial Analysis is internal-only. Sample sizes and confidence intervals are shown when we have them. There is no “#1.”
Live snapshot from the public Model Intel publication (slug current), regenerated at most hourly.
Jump a class, or read them in order. Coding rows always name the harness. n is the first-party sample; intervals are shown when present.
Multi-file repository work: implement, recover, verify. Ranked as (harness, model, role).
community
External/community fusion for code.agentic
No attributed evidence published for this class yet.
Generated websites, components, and full-stack UI. Taste plus deployable quality.
first_party
First-party n=129 dominates
n=129 first-party success 1.000
Finding real defects without comment fatigue. Cross-family review is required.
insufficient
External/community fusion for code.review
n=8 first-party success 1.000
Strategic, second-order, long-horizon analysis.
Awaiting first publication for this class.
No attributed evidence published for this class yet.
Multi-step tool calling with schema fidelity and recovery.
first_party
First-party n=1108 dominates
n=1108 first-party success 0.940
User-visible prose, brand voice, and durable narratives.
Awaiting first publication for this class.
No attributed evidence published for this class yet.
Schema-valid JSON / Output.object without silent field loss.
first_party
First-party n=569 dominates
n=569 first-party success 0.916
Documents, transcripts, and large codebases. Effective window, not marketed size.
Awaiting first publication for this class.
No attributed evidence published for this class yet.
PDF, OCR, diagram, and image understanding.
Awaiting first publication for this class.
No attributed evidence published for this class yet.
Binary and cheap structured gates. Recoverable if wrong.
Awaiting first publication for this class.
No attributed evidence published for this class yet.
Native X search and community recency. Discovery, not deciding.
first_party
First-party n=165 dominates
n=165 first-party success 0.952
Work classes are the spine. External benchmarks, community consensus, and HiveBase first-party outcomes all hang off the same eleven jobs so they stay commensurable. Fusion never promotes a default from community recency alone — that channel is capped and decays.
Coding is harness-keyed. The ranked unit is (harness, model, role). An executor on Claude Code is not comparable to the same weights inside Cursor or Codex. Review must be cross-family.
First-party needs a sample. HiveBase treats n < 20 as below the first-party threshold. Those classes can still carry a routing opinion; they are marked, not dressed up as settled measurement.
Artificial Analysis is not rehosted. AA indexes inform internal routing. This page will only say they are tracked and point at artificialanalysis.ai. Attributed sources (Terminal-Bench, SWE-Rebench, Design Arena, LMArena, BFCL, Epoch) appear as name + outbound link.
Saturated suites (SWE-bench Verified, HumanEval, retired open LLM leaderboards) are weight-zero. A canary that fails the quality floor auto-reverts; those events are public when the live feed includes them.
Promotions that did not hold stay on the record. Auto-reverts are not hidden behind a successful later default.
No public changelog entries in this snapshot.
Ingest health for this snapshot. A skipped or empty refresh is shown, not implied successful. Artificial Analysis appears here as a tracked source, never as a copied score.
| Source | Finished | Status |
|---|---|---|
| fusion | Aug 18, 2026, 06:00 AM UTC | ok |
| overlay | Aug 18, 2026, 06:00 AM UTC | skipped |
| registry | Aug 18, 2026, 06:00 AM UTC | ok |
| grok_release_radar | Aug 18, 2026, 06:00 AM UTC | error |
| grok_consensus | Aug 18, 2026, 05:30 AM UTC | ok |
| registry | Aug 18, 2026, 05:00 AM UTC | ok |
| overlay | Aug 18, 2026, 05:00 AM UTC | skipped |
| openrouter | Aug 18, 2026, 05:00 AM UTC | skipped |
| first_party | Aug 18, 2026, 05:00 AM UTC | ok |
| wulong_arena | Aug 18, 2026, 05:00 AM UTC | ok |
| hf_hub | Aug 18, 2026, 05:00 AM UTC | skipped |
| lmarena_text | Aug 18, 2026, 05:00 AM UTC | error |
| artificial_analysisTracked internally — artificialanalysis.ai | Aug 18, 2026, 05:00 AM UTC | skipped |
| overlay | Aug 18, 2026, 04:00 AM UTC | skipped |
| registry | Aug 18, 2026, 04:00 AM UTC | ok |
| overlay | Aug 18, 2026, 03:00 AM UTC | skipped |
| registry | Aug 18, 2026, 03:00 AM UTC | ok |
| overlay | Aug 18, 2026, 02:00 AM UTC | skipped |
| registry | Aug 18, 2026, 02:00 AM UTC | ok |
| overlay | Aug 18, 2026, 01:00 AM UTC | skipped |
| registry | Aug 18, 2026, 01:00 AM UTC | ok |
| overlay | Aug 18, 2026, 12:00 AM UTC | skipped |
| registry | Aug 18, 2026, 12:00 AM UTC | ok |
| grok_release_radar | Aug 18, 2026, 12:00 AM UTC | ok |
| overlay | Aug 17, 2026, 11:00 PM UTC | skipped |
| registry | Aug 17, 2026, 11:00 PM UTC | ok |
| overlay | Aug 17, 2026, 10:00 PM UTC | skipped |
| registry | Aug 17, 2026, 10:00 PM UTC | ok |
| overlay | Aug 17, 2026, 09:00 PM UTC | skipped |
| registry | Aug 17, 2026, 09:00 PM UTC | ok |
| overlay | Aug 17, 2026, 08:00 PM UTC | skipped |
| registry | Aug 17, 2026, 08:00 PM UTC | ok |
| overlay | Aug 17, 2026, 07:00 PM UTC | skipped |
| registry | Aug 17, 2026, 07:00 PM UTC | ok |
| overlay | Aug 17, 2026, 05:00 PM UTC | skipped |
| registry | Aug 17, 2026, 05:00 PM UTC | ok |
| overlay | Aug 17, 2026, 04:00 PM UTC | skipped |
| registry | Aug 17, 2026, 04:00 PM UTC | ok |
| registry | Aug 17, 2026, 03:00 PM UTC | ok |
| overlay | Aug 17, 2026, 03:00 PM UTC | skipped |
| overlay | Aug 17, 2026, 02:00 PM UTC | skipped |
| registry | Aug 17, 2026, 02:00 PM UTC | ok |
| overlay | Aug 17, 2026, 01:00 PM UTC | skipped |
| registry | Aug 17, 2026, 01:00 PM UTC | ok |
| overlay | Aug 17, 2026, 12:00 PM UTC | skipped |
| registry | Aug 17, 2026, 12:00 PM UTC | ok |
| grok_release_radar | Aug 17, 2026, 12:00 PM UTC | ok |
| overlay | Aug 17, 2026, 11:00 AM UTC | skipped |
| registry | Aug 17, 2026, 11:00 AM UTC | ok |
| overlay | Aug 17, 2026, 10:00 AM UTC | skipped |
| registry | Aug 17, 2026, 10:00 AM UTC | ok |
| overlay | Aug 17, 2026, 09:00 AM UTC | skipped |
| registry | Aug 17, 2026, 09:00 AM UTC | ok |
| overlay | Aug 17, 2026, 08:00 AM UTC | skipped |
| registry | Aug 17, 2026, 08:00 AM UTC | ok |
| registry | Aug 17, 2026, 07:00 AM UTC | ok |
| overlay | Aug 17, 2026, 07:00 AM UTC | skipped |
| publication | Aug 17, 2026, 06:30 AM UTC | ok |
| fusion | Aug 17, 2026, 06:00 AM UTC | ok |
| grok_release_radar | Aug 17, 2026, 06:00 AM UTC | ok |
| registry | Aug 17, 2026, 06:00 AM UTC | ok |
| overlay | Aug 17, 2026, 06:00 AM UTC | skipped |
| grok_consensus | Aug 17, 2026, 05:30 AM UTC | ok |
| hf_hub | Aug 17, 2026, 05:00 AM UTC | skipped |
| first_party | Aug 17, 2026, 05:00 AM UTC | ok |
| wulong_arena | Aug 17, 2026, 05:00 AM UTC | ok |
| openrouter | Aug 17, 2026, 05:00 AM UTC | skipped |
| artificial_analysisTracked internally — artificialanalysis.ai | Aug 17, 2026, 05:00 AM UTC | skipped |
| lmarena_text | Aug 17, 2026, 05:00 AM UTC | error |
| registry | Aug 17, 2026, 05:00 AM UTC | ok |
| overlay | Aug 17, 2026, 05:00 AM UTC | skipped |
| overlay | Aug 17, 2026, 04:00 AM UTC | skipped |
| registry | Aug 17, 2026, 04:00 AM UTC | ok |
| registry | Aug 17, 2026, 03:00 AM UTC | ok |
| overlay | Aug 17, 2026, 03:00 AM UTC | skipped |
| overlay | Aug 17, 2026, 02:00 AM UTC | skipped |
| registry | Aug 17, 2026, 02:00 AM UTC | ok |
| overlay | Aug 17, 2026, 01:00 AM UTC | skipped |
| registry | Aug 17, 2026, 01:00 AM UTC | ok |
| overlay | Aug 17, 2026, 12:00 AM UTC | skipped |
| registry | Aug 17, 2026, 12:00 AM UTC | ok |
| grok_release_radar | Aug 17, 2026, 12:00 AM UTC | ok |
| overlay | Aug 16, 2026, 11:00 PM UTC | skipped |
| registry | Aug 16, 2026, 11:00 PM UTC | ok |
| overlay | Aug 16, 2026, 10:00 PM UTC | skipped |
| registry | Aug 16, 2026, 10:00 PM UTC | ok |
| overlay | Aug 16, 2026, 09:00 PM UTC | skipped |
| registry | Aug 16, 2026, 09:00 PM UTC | ok |
| overlay | Aug 16, 2026, 08:00 PM UTC | skipped |
| registry | Aug 16, 2026, 08:00 PM UTC | ok |
| registry | Aug 16, 2026, 07:00 PM UTC | ok |
| overlay | Aug 16, 2026, 07:00 PM UTC | skipped |
| grok_release_radar | Aug 16, 2026, 06:00 PM UTC | ok |
| overlay | Aug 16, 2026, 06:00 PM UTC | skipped |
| registry | Aug 16, 2026, 06:00 PM UTC | ok |
| overlay | Aug 16, 2026, 05:00 PM UTC | skipped |
| registry | Aug 16, 2026, 05:00 PM UTC | ok |
| overlay | Aug 16, 2026, 04:00 PM UTC | skipped |
| registry | Aug 16, 2026, 04:00 PM UTC | ok |
| overlay | Aug 16, 2026, 03:00 PM UTC | skipped |
| registry | Aug 16, 2026, 03:00 PM UTC | ok |
| registry | Aug 16, 2026, 02:00 PM UTC | ok |
| overlay | Aug 16, 2026, 02:00 PM UTC | skipped |
| registry | Aug 16, 2026, 01:00 PM UTC | ok |
| overlay | Aug 16, 2026, 01:00 PM UTC | skipped |
| overlay | Aug 16, 2026, 12:00 PM UTC | skipped |
| registry | Aug 16, 2026, 12:00 PM UTC | ok |
| grok_release_radar | Aug 16, 2026, 12:00 PM UTC | ok |
| overlay | Aug 16, 2026, 11:00 AM UTC | skipped |
| registry | Aug 16, 2026, 11:00 AM UTC | ok |
| overlay | Aug 16, 2026, 10:00 AM UTC | skipped |
| registry | Aug 16, 2026, 10:00 AM UTC | ok |
| overlay | Aug 16, 2026, 09:00 AM UTC | skipped |
| registry | Aug 16, 2026, 09:00 AM UTC | ok |
| overlay | Aug 16, 2026, 08:00 AM UTC | skipped |
| registry | Aug 16, 2026, 08:00 AM UTC | ok |
| registry | Aug 16, 2026, 07:00 AM UTC | ok |
| overlay | Aug 16, 2026, 07:00 AM UTC | skipped |
| publication | Aug 16, 2026, 06:30 AM UTC | error |
| registry | Aug 16, 2026, 06:00 AM UTC | ok |
| overlay | Aug 16, 2026, 06:00 AM UTC | skipped |
| grok_release_radar | Aug 16, 2026, 06:00 AM UTC | ok |
| grok_consensus | Aug 16, 2026, 05:30 AM UTC | ok |
| hf_hub | Aug 16, 2026, 05:00 AM UTC | skipped |
| first_party | Aug 16, 2026, 05:00 AM UTC | ok |
| wulong_arena | Aug 16, 2026, 05:00 AM UTC | ok |
| openrouter | Aug 16, 2026, 05:00 AM UTC | skipped |
| lmarena_text | Aug 16, 2026, 05:00 AM UTC | error |
| artificial_analysisTracked internally — artificialanalysis.ai | Aug 16, 2026, 05:00 AM UTC | skipped |
| overlay | Aug 16, 2026, 05:00 AM UTC | skipped |
| registry | Aug 16, 2026, 05:00 AM UTC | ok |
| overlay | Aug 16, 2026, 05:00 AM UTC | skipped |
| registry | Aug 16, 2026, 05:00 AM UTC | ok |
| overlay | Aug 16, 2026, 04:00 AM UTC | skipped |
| registry | Aug 16, 2026, 04:00 AM UTC | ok |
| registry | Aug 16, 2026, 03:00 AM UTC | ok |
| overlay | Aug 16, 2026, 03:00 AM UTC | skipped |
A model that is right for cheap classification is the wrong unit for multi-file coding. HiveBase publishes a verdict per work class. For code.agentic, code.frontend, and code.review the unit is (harness, model, role) — the same model on Claude Code versus Cursor is not the same system.
Artificial Analysis is tracked internally for routing. Their terms do not allow us to rehost numeric scores. When AA is a source, the page says so and links to artificialanalysis.ai.
If a canary drops below HiveBase's quality floor, routing returns to the previous default. Public auto-reverts are listed in the changelog so a promotion that did not hold is visible, not buried.
When the public snapshot is reachable, the receipt reads live. If environment, table, or query fail, this page renders a labeled preview snapshot so the marketing site never depends on production being up.
HiveBase uses these work-class verdicts when it dispatches coding missions and the rest of the company. The same boundaries, receipts, and undo as everywhere else.