Clean Separation Between Active Rate Cards and Archive Pages
AI engineering teams routinely encounter blog posts, tutorials, and older codebase configs referencing model IDs that providers have superseded. Rather than silently deleting legacy IDs (which breaks engineering documentation and historical cost audits), TokenCalc AI flags retired entries with explicit status: "retired" metadata and a successor field.
Calculators, smart routing estimators, and comparison dropdowns draw strictly from active models, while individual archive pages stay accessible with migration warnings.
| Retired Model | Provider | Recommended Active Successor | Successor Input / Output |
|---|---|---|---|
| OpenAI o1 | OpenAI | GPT-6 Astra | $10.00 / $50.00 |
| OpenAI o3-mini | OpenAI | GPT-6.1 Sol | $2.00 / $10.00 |
| Claude 3.5 / 3.7 Sonnet | Anthropic | Claude Sonnet 5.5 | $2.00 / $10.00 |
| Claude 3.5 Haiku | Anthropic | Claude Haiku 5.5 | $0.10 / $0.50 |
| Gemini 2.0 Flash / 1.5 Flash | Gemini 3.5 Flash-Lite | $0.30 / $2.50 | |
| Gemini 1.5 Pro | Gemini 3.1 Pro (Preview) | $2.00 / $12.00 | |
| DeepSeek V3 | DeepSeek | DeepSeek Flash | $0.15 / $0.60 |
| DeepSeek R1 | DeepSeek | DeepSeek V4 Pro | $0.66 / $1.98 |
| Grok 2 | xAI | Grok 4.7 | $2.00 / $6.00 |
Retired model successor mapping in data/models.ts
Redirects & Editorial Continuity
For high-traffic legacy comparison intents like Claude 3.5 Sonnet alternatives, permanent 308/301 routing points directly to updated Claude Sonnet 5.5 coverage so links from existing engineering docs land cleanly on current economics.