On this index. Eight profiles are published from the 21-model dataset. Scores are copied from the comparison JSON, last verified 2026-06-26. Open a profile for that model's task, cost and privacy story, or use the match engine to choose from all 21. This is not full LOCK-05 coverage of every model id.
These profiles are entity pages for models that already exist in the published comparison dataset. Each URL is one model id, so a business can cite GPT-5.4, Claude Opus 4.8 or DeepSeek V3 without landing on a generic roundup. The index exists so the profiles are not orphans. It does not rescore them and it does not add models that are absent from the JSON. Wave 1 shipped GPT-5.4 (weighted 85), Claude Sonnet 4.6 (87) and Gemini 3 Flash (86). Wave 2 adds GPT-4o (84), Claude Opus 4.8 (81), DeepSeek V3 (80), Grok 4.1 (78) and Mistral Large 3 (83). All figures were last verified on 2026-06-26 under scoring version v1.0. Thirteen models still have no profile page. Use a profile when you need the task, cost and privacy story for that model. Use the match engine when the job, the budget and the privacy constraint should choose the model for you. Affiliate links are off.
Profiles in this set
| Profile | Weighted | Truth | Decision the page is for |
|---|---|---|---|
| Claude Sonnet 4.6 | 87 | 96 | Truth and privacy against the workhorse price |
| Gemini 3 Flash | 86 | 80 | Speed, million-token context and support-volume cost |
| GPT-5.4 | 85 | 82 | Task fit, API cost and privacy on the OpenAI workhorse |
| GPT-4o | 84 | 80 | Speed, voice and multimodal jobs on the OpenAI rate card |
| Mistral Large 3 | 83 | 80 | EU sovereignty, privacy 92 and self-host options |
| Claude Opus 4.8 | 81 | 92 | High-stakes analysis at the frontier rate card |
| DeepSeek V3 | 80 | 68 | Budget API cost versus China-jurisdiction privacy flags |
| Grok 4.1 | 78 | 79 | Live news and social research with compliance flags false |
Figures from /data/ai-comparison-2026.json. Last verified 2026-06-26. Scoring version v1.0. Affiliate links are off.
How a profile differs from the comparison table
The homepage table ranks the full set. A profile keeps one model id stable. Privacy and cost questions that use these rows include enterprise AI privacy cost, when a cheaper model is safer and RAG for legal work. Factor definitions stay on methodology.
Not sure which profile to open? Start with the match engine. It is the decision tool. These pages explain the rows it can choose.
FAQ
How many models have profiles?
Eight profiles are published. The dataset still holds 21 models. The remaining models stay on the homepage table until a profile exists. This is not LOCK-05 complete coverage of all 21.
Do profile pages use different scores from the JSON?
No. Every figure is copied from ai-comparison-2026.json, last verified 2026-06-26 under scoring version v1.0.
Are these paid placements?
No. Affiliate links are off. A profile does not change a model's score.