Anthropic API pricing, price hikes and cheaper alternatives
Anthropic API (anthropic.com) is a LLM API. Pay as you go per token, with a separate input and output price for each model, plus usage-based add-ons such as Web Search, Code Execution and Managed Agents.. No free plan.
What Anthropic API costs
- Top model: $200/mo (Fable 5.1 at $10/MTok input and $50/MTok output: 10 x $10 + 2 x $50 = $200/mo. LMArena rank not verified (the leaderboard returned HTTP 403), so Fable 5.1 is taken as the most expensive general-purpose model; Opus 5.5 would give $80 if it ranks higher.)
- Cheapest model: $2/mo (Haiku 5.5 at $0.10/MTok input and $0.50/MTok output (the up-to-100K-prompt tier, which covers a 1,000-token prompt): 10 x $0.10 + 2 x $0.50 = $2/mo. Cheapest current model on the official page; tool use and a 128K context window are not stated per model there, so it is assumed to qualify.)
- gpt-oss-120b: not offered (n/a: Anthropic does not sell gpt-oss-120b per token.)
Plan compared: Fable 5.1 pay-as-you-go at standard rates: 10M input and 2M output tokens a month.. Monthly USD cost of one reference workload: 10,000 requests a month, totaling 10M input and 2M output tokens, at standard pay-as-you-go rates with no caching or batch discounts. Top model is the product's highest-ranked model on the LMArena Text leaderboard. Cheapest model is its lowest-cost model with tool calling and a 128K context. gpt-oss-120b compares hosts on the same open model.
What Anthropic API does
Core: Frontier-grade models, 1M-token context, Hosted agent tools, Prompt caching discounts.
Important: Image and PDF input, Batch and flex pricing, No-training, zero retention (partly), Compliance and data residency (partly).
Also has: OpenAI-compatible endpoint (partly), Dedicated model endpoints (partly), Flat-rate coding plan (partly).
Anthropic API price hikes
No price hikes on record for Anthropic API.
Cheaper alternatives to Anthropic API
At 12M tokens a month on its top model, Anthropic API costs $200/mo. 1 LLM API costs less:
- Gemini API: $40/mo (Pay-as-you-go Standard paid tier, text only, list prices as of 2026-10-17), covers 96% of Anthropic API; missing or partial: Dedicated model endpoints
Anthropic API alternatives: 3 LLM APIs compared, at every size.
Other kinds of products that cover part of Anthropic API
- Perplexity (AI assistants): covers 54% of Anthropic API; missing or partial: 1M-token context (partly), Prompt caching discounts, Image and PDF input, Batch and flex pricing, Dedicated model endpoints
- Hugging Face (GPU cloud & model hosting): covers 35% of Anthropic API; missing or partial: Frontier-grade models (partly), 1M-token context (partly), Hosted agent tools, Prompt caching discounts, Image and PDF input (partly)
Every change we logged
- Oct 12, 2026 · Price cut: Sonnet 5.5 cache-read price was halved, from $0.20 to $0.10 per million tokens.
- Oct 12, 2026 · Price cut: Haiku 5.5 launched at $0.10 input and $0.50 output per million tokens for prompts up to 100K tokens, with $0.50 and $2.50 above that. Haiku 4.5 moved to Legacy.
- Sep 28, 2026 · Price cut: Opus 5.5 became the flagship at $4 input and $20 output per million tokens, below Opus 5's $5 and $25. Opus 5 is now Legacy.
- Jul 6, 2026 · Price cut: Sonnet 5 replaced Sonnet 4.6 at intro rates of $2 input and $10 output per million tokens, below $3 and $15. Standard $3/$15 was to resume after the intro period ended Aug 31, 2026.
- Mar 16, 2026 · Price cut: Opus 4.6 and Sonnet 4.6 lost their above-200K-token tiers. Above 200K, Opus input fell from $10 to $5 and output from $37.50 to $25; Sonnet input fell from $6 to $3 and output from $22.50 to $15. Caching moved to single rates.
Sources
Pricing page: claude.com/pricing. Prices last checked Oct 16, 2026. Prices come from Anthropic's official pricing page (read 2026-10-17) and the PricingSaaS change log, a third-party tracker not checked against Anthropic. LMArena was blocked (HTTP 403) and the web-search budget was used up, so the top-model choice is unverified and the hike check rests on the PricingSaaS log alone.