Anthropic is an AI research company focused on ensuring that advanced AI systems are safe, interpretable, and aligned with human values.
Pricing Change History
15 changes · 3 summariesNew Fable 5.2 model replaces Fable 5.1 as the flagship API model, priced 20% lower: input $8, output $40, cache read $0.20, cache write $10 per MTok (vs. Fable 5.1 at $10/$50/$0.25/$12.50). Fable 5.1 stays available under Legacy models at unchanged rates; fast mode (2x standard pricing) now also covers Fable 5.2.
Haiku 5.5 replaces Haiku 4.5 at lower rates ($0.10/$0.50 per MTok in/out, tiered at 100K prompts); Haiku 4.5 moved to Legacy. Sonnet 5.5 cache read $0.20 to $0.10 per MTok.
Includes Oct 8 · first seen Oct 8, 2026
New flagship model Opus 5.5 (“Daily driver for agentic coding and enterprise work”) launched at $4/MTok input, $20/MTok output, replacing Opus 5; Opus 5 moved to Legacy models at its original $5/$25 pricing.
Opus 4.8 tagline changed from ‘Most intelligent model for agents and coding’ to ‘Ideal for complex agentic coding and enterprise work’, emphasizing enterprise use.
Opus 4.8 launched as new flagship API model (same $5/$25 per MTok pricing as Opus 4.7); Opus 4.7 moved to legacy section. New fast mode available for Opus 4.8 at 2x standard pricing for up to 2.5x faster speeds.
Pro plan Microsoft Office integrations consolidated: Claude for Excel, PowerPoint, and Word (beta) replaced by Claude for Microsoft 365 and Claude for Microsoft Outlook.
Managed Agents tool added to Claude Platform features: build and deploy agents at scale with composable APIs at $0.08 per session-hour for active runtime. API tools section relabeled from ‘Pricing for tools’ to ‘Pricing for Claude Platform features’, reflecting expanded platform positioning beyond standalone tools.
Sonnet 4.6 drops the 200K-token tiered pricing used by Sonnet 4.5. Flat $3/$15 MTok replaces tiers of $3/$6 input and $15/$22.50 output. Extended caching tiers ($7.50/$0.60) also removed.
Includes 5 weeklies · first seen Jan 5, 2026
API pricing simplified by removing volume-based tiered pricing [Opus 4.6: flat $5/MTok input (was $5 for ≤200K, $10 for >200K tokens), $25/MTok output (was $25 for ≤200K, $37.50 for >200K); Sonnet 4.6: flat $3/MTok input (was $3 for ≤200K, $6 for >200K), $15/MTok output (was $15 for ≤200K, $22.50 for >200K); all prompt caching simplified to single rates; Haiku 4.5 unchanged]
Sonnet 4.5 retired from current models and moved to legacy section; now listed as first legacy model (previously Opus 4.5 was first legacy) Sonnet 4.6 introduced to current API models replacing Sonnet 4.5 at same price point ($3 input / $15 output per MTok)
Opus 4.5 replaced by Opus 4.6 with tiered context-length pricing: input now $5/MTok (<=200K) or $10/MTok (>200K); output $25/MTok (<=200K) or $37.50/MTok (>200K); prompt caching adds >200K tier.
Opus 4.5 moved from Latest models to Legacy models section; Opus 4.6 input pricing: added >200K token tier at $10/MTok (2x the ≤200K rate of $5/MTok); Opus 4.6 output pricing: added >200K token tier at $37.50/MTok (1.5x the ≤200K rate of $25/MTok); Opus 4.6 prompt caching: added >200K token tier (Write $12.50/MTok, Read $1/MTok)
Holiday promotional banner removed: paid subscribers no longer receive 2x higher usage limits. The temporary ‘Happy holidays’ promotion offering doubled usage through 12/31 has expired.
Current plans
| Plan | Price | |
|---|---|---|
| Opus 5.5 | Usage-based | |
Also priced by usage
| ||
| Sonnet 5.5 | Usage-based | |
Also priced by usage
| ||
| Opus 4.5 | Usage-based | |
Also priced by usage
| ||
| Haiku 5.5 | Usage-based | |
Also priced by usage
| ||
| Sonnet 4.5 | Usage-based | |
Also priced by usage
| ||
| Opus 4.6 | Usage-based | |
Also priced by usage
| ||
| Opus 4.7 | Usage-based | |
Also priced by usage
| ||
| Sonnet 4.6 | Usage-based | |
Also priced by usage
| ||
| Fable 5.1 | Usage-based | |
Also priced by usage
| ||
| Opus 4.8 | Usage-based | |
Also priced by usage
| ||
| Fable 5 | Usage-based | |
Also priced by usage
| ||
| Haiku 4.5 | Usage-based | |
Also priced by usage
| ||
| Sonnet 5 | Usage-based | |
Also priced by usage
| ||
| Opus 5 | Usage-based | |
Also priced by usage
| ||
| Fable 5.2 | Usage-based | |
Also priced by usage
| ||
| Code ExecutionAdd-on | $0.05 per container hour | |
| Web SearchAdd-on | $10 per 1,000 Search Calls | |
| Managed AgentsAdd-on | $0.08 per session‑hour | |
Company Facts
Pulse news about Anthropic
View all →




How Anthropic compares with its peers
| Company | Entry price |
|---|---|
O OpenAILast change: Nov 13, 2026![]() | $8 /mo |
G GoogleLast change: Oct 12, 2026 | Contact |
M MicrosoftLast change: Nov 16, 2026![]() | $5 /user/mobilled annually |
X XLast change: Nov 9, 2026 | $0.50 per 1M input tokens (cached) + $6 per 1M output tokens |
M Mistral AILast change: Nov 21, 2026![]() | $14.99 /mo |
S ScaleLast change: Aug 28, 2026![]() | Free |
Anthropic pricing FAQ
- How much does Anthropic cost?
- Anthropic prices on usage: the Opus 5.5 plan is $0.20 per MTok cache read + $4 per MTok input.
- Does Anthropic have a free plan?
- No. Anthropic does not list a free tier.
- What pricing model does Anthropic use?
- Anthropic uses usage-based pricing across 15 plans.
- When did Anthropic last change its pricing?
- Nov 16, 2026: Anthropic API launches Fable 5.2 at 20% lower prices, replacing Fable 5.1.
- How often has Anthropic changed its pricing?
- Pulse has recorded 15 pricing changes for Anthropic since it started tracking on Nov 17, 2024; the pricing page was last checked on Nov 16, 2026.
- What does “tracked since” mean?
- It's the date of the first saved copy of the pricing page that Pulse checked for changes against a later copy. For a few companies we hold older copies that Pulse never checked for changes, and those don't change this date. If Pulse has never checked two copies, it's the date of the first saved copy.



