Claude models in 2026: Fable 5.1, Opus 5, Sonnet 5, Haiku 4.5, and legacy models explained
Updated: Sep 17
Anthropic's current Claude lineup in September 2026 is Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, and Claude Haiku 4.5. Older 4.x models can still be active in the API, but Anthropic now labels many of them legacy and recommends migration to the 5-series replacements.
The four current models occupy different cost, latency, and reasoning positions. Fable 5.1 is the highest-end long-horizon model, Opus 5 is Anthropic's recommended starting point for most demanding workloads, Sonnet 5 combines speed and intelligence at a lower price, and Haiku 4.5 remains the fastest low-cost option.
··········
THE CURRENT CLAUDE MODEL MAP
Anthropic's official model overview now presents Fable 5.1, Opus 5, Sonnet 5, and Haiku 4.5 as the current lineup. All four accept text and image input and produce text output, but their context, reasoning system, price, and latency differ.
........
Current model | Context / max output | API price / MTok | Positioning |
|---|---|---|---|
Claude Fable 5.1 | 1M / 128K | $10 input / $50 output | Demanding reasoning and long-horizon agents |
Claude Opus 5 | 1M / 128K | $5 / $25 | Complex agentic coding and enterprise work |
Claude Sonnet 5 | 1M / 128K | $2 / $10 | Fast, high-capability general and agentic work |
Claude Haiku 4.5 | 200K / 64K | $1 / $5 | Fastest and lowest-cost current model |
........
··········
FABLE 5.1 IS THE HIGH-END LONG-HORIZON MODEL
Claude Fable 5.1 has a 1M-token context window, 128K maximum output, always-on adaptive thinking, and a June 2026 reliable knowledge cutoff. Anthropic recommends it for demanding reasoning and long-horizon agentic work when Opus 5 at higher effort is still insufficient.
··········
OPUS 5 IS THE RECOMMENDED STARTING POINT FOR MOST HARD WORK
Anthropic's current model guide says to start with Opus 5 for most workloads. It shares the 1M context and 128K output ceilings of Fable 5.1 at half the base input and output price.
··········
SONNET 5 CHANGED THE PRICE-PERFORMANCE MIDDLE
Sonnet 5 launched at $2 per million input tokens and $10 per million output tokens, and Anthropic later made that introductory price permanent. It also has a 1M context window and 128K output, making it cheaper than legacy Sonnet 4.6 while remaining the current Sonnet generation.
··········
HAIKU 4.5 REMAINS THE CURRENT FASTEST LOW-COST MODEL
Haiku 4.5 remains active in the current lineup with a 200K context window, 64K maximum output, extended thinking, and $1/$5 API pricing. It is the low-latency choice rather than a direct substitute for Fable or Opus reasoning depth.
··········
LEGACY DOES NOT MEAN RETIRED
Anthropic marks models such as Sonnet 4.6, Opus 4.6, Sonnet 4.5, and Opus 4.5 as legacy while they remain active. Separately, older models such as Sonnet 4, Opus 4, Opus 4.1, Sonnet 3.7, Haiku 3.5, and Haiku 3 have already reached API retirement dates.
··········
DS COST NORMALIZATION
For a workload with 10M uncached input tokens and 2M output tokens, base token charges are about $20 on Haiku 4.5 (10×$1 + 2×$5), $40 on Sonnet 5 (10×$2 + 2×$10), $100 on Opus 5 (10×$5 + 2×$25), and $200 on Fable 5.1 (10×$10 + 2×$50). This comparison excludes prompt caching, batch discounts, and regional pricing adjustments.
··········
MODEL NAMES, PLAN ACCESS, AND API BILLING ARE SEPARATE QUESTIONS
A model can be current in the Claude API without being included in every Claude subscription allowance. Claude paid subscriptions and Claude API/Console usage are billed separately, and high-end Fable access can depend on plan type or usage credits.
··········
FOLLOW US FOR MORE
··········
DATA STUDIOS
··········
