Sources & coverage

Official pages checked 30 Sep 2026. Prices are dated source snapshots, not a live billing feed or a guarantee of purchasing eligibility.

What is included

The calculator also includes 113 text-model quotes imported from the OpenRouter catalogon 2026-09-30. These are explicitly labelled OpenRouter channel prices, separate from direct-vendor quotes. Search by model or brand. Unsupported conditional rates are excluded.

Eight Chinese text models, eleven international API comparison options, nine media quotes and five subscription plans. Text calculations compare the same token counts; models can produce different outputs. USD is the original quote currency. Other currencies use European Central Bank reference rates dated 2026-09-30, crossed from EUR to USD. These are informational conversions, not card or transaction rates.ECB source ↗

What is held out

• Qwen3.8 cache prices: the vendor requires a console quote. Base input and output estimates work; nonzero cached usage stops the estimate.
• Kimi memberships: the help center lists legacy tiers while newer documentation mentions Go / Plus / Pro / Max / Ultra. Prices are held out until purchasing eligibility and current tiers agree.
• Guaranteed monthly subscription token totals: weekly and rolling allowances cannot establish these. GLM now has a separate four-quota-week scenario estimate using published credit conversions. MiniMax’s undisclosed numerical quota remains outside token/value rankings.
• Tax, cache writes/storage, tools, retry and input-material charges are not silently included. Enter known extra fees in the text calculator.

Source register

We watch OpenRouter rankings for platform usage trends and Artificial Analysis for benchmark context. Neither is a whole-market popularity census. These sources guide the shortlist; prices come from official vendor pages. Qwen quotes use Singapore International deployment.

• OpenRouter model catalog API ↗

  Automatically imported OpenRouter channel quotes, separate from direct vendor prices. Text-token estimates only; unsupported conditional quotes are excluded.

• OpenAI API pricing ↗

  Standard API short/long-context brackets, not Batch/Flex/Fast. Regional uplifts and cache writes excluded.

• GPT-6.1 Sol API ↗

  Official API model identifier and base rate.

• GPT-6 Astra API ↗

  Official API model identifier and base rate.

• GPT-6 Luna API ↗

  Current Luna family API option.

• Claude API pricing ↗

  Standard global Claude API pricing; cache writes and tools separate.

• Claude Opus 5.5 ↗

  API access and model specifications.

• Claude Sonnet 5.5 ↗

  API access and model specifications.

• Claude Haiku 4.5 ↗

  Current Haiku family model, not the newest Claude release overall.

• Gemini Developer API pricing ↗

  Standard paid tier. Flash 3.8 current offer expires 31 Dec 2026; storage and grounding excluded.

• Grok API pricing ↗

  Global endpoint; long-context bracket begins at 200K prompt tokens.

• Grok 4.7 API ↗

  Current flagship API model.

• Qwen official API pricing ↗

  Singapore International deployment. No free quota or regional substitution in estimates.

• Qwen cache pricing rules ↗

  Qwen3.8 cache rates are exceptions to generic discounts and require console quotes; unknown here.

• OpenRouter usage rankings ↗

  Candidate discovery from platform token traffic; not market-wide users, spending or quality. Check weekly.

• Artificial Analysis model leaderboard ↗

  Independent benchmark reference for candidate review. Do not substitute blended leaderboard prices for vendor quotes.

• DeepSeek API pricing ↗

  Billing documentation takes precedence over the older platform marketing table. Peak/off-peak rates are separate.

• Z.AI API pricing ↗

  USD pay-as-you-go rates. Cache storage is currently a limited-time free item; not assumed permanently free.

• GLM-5.3 model guide ↗

  Text input, 1M context; reasoning enabled.

• GLM-5.3-Flash model guide ↗

  Multimodal input, text output, 1M context.

• MiniMax API pricing ↗

  USD list. M2.7 cache writes have a separate price. Video prices below cover output; input materials may cost extra.

• MiniMax model catalog ↗

  M2.7 remains listed; M3 request-length brackets are selected explicitly in the calculator.

• MiniMax M Plan — monthly ↗

  Browser monthly view: Go $22, Explore $55, Build $132 recurring. First-month 50% offer not used. Supersedes old Token Plan search snippets.

• GLM Coding Plan — subscription page ↗

  Browser Monthly selection: Lite $18, Pro $80, Max $168. Annual equivalents excluded.

• GLM Coding Plan — quota rules ↗

  Weekly and rolling five-hour credits both apply. Published per-token multipliers support a model-specific four-week example, not a guaranteed monthly total. Temporary promotions and MCP charges are excluded.

• Kimi Code membership guide ↗

  Tool introduction only. Legacy membership pricing conflicts with newer tier names; Kimi prices held out.

• DeepSeek Harness Web UI guide ↗

  Official API documentation links to this guide. Developer preview, not an independent performance review.

• MiniMax M3 announcement ↗

  Provider announcement; performance claims are not independently tested.

Review and freshness

Checks record observed prices and terms, not a guarantee of future availability. The update process reads a fixed catalog and watches registered official sources for changes. Direct-vendor snapshots keep their own review dates. Failed checks retain the last snapshot and never advance its review date. Dated snapshots remain available for comparison without an automatic expiry. Check the linked provider before purchasing. A successful download is not a review of changed billing rules.