Latest7 Oct 2026Small/fast LLM
Claude Haiku 5.5
Anthropic's new small, low-cost model in the Claude 5.5 family, meant for high-volume, speed-sensitive work such as customer support and browser use, and as a helper to larger Claude models. Anthropic says it costs around 75% less to run than before. It is the first Haiku-class model with an adjustable effort setting.
- 1M-token context window; 128K max output tokens
- Adjustable effort setting, a first for Haiku-class models
- Text and image input, tool use, computer and browser use
- OSWorld 2.1 (offline subset) 72.4% (Anthropic)
- Humanity's Last Exam 45.9% without tools, 57.4% with tools (Anthropic)
AccessClaude API (claude-haiku-5-5), Amazon Bedrock, Google Cloud and Microsoft Foundry; Anthropic says it is 'available now on all platforms'
Price$0.10 input / $0.50 output per million tokens (prompts up to 100K tokens); $0.50 / $2.50 for longer prompts
28 Sep 2026Mid-size LLM
Claude Sonnet 5.5
The second model in Anthropic's Claude 5.5 family and the mid-priced option below Opus 5.5. Anthropic calls it a clear upgrade over Claude Sonnet 5 that runs more than 30% faster and costs up to 30% less for most work, aimed at everyday tasks, debugging and document creation.
- 1M-token context window, 128K max output (docs)
- Terminal-Bench 4.0: 70.6% (Anthropic-reported)
- Knowledge cutoff June 2026 (docs)
- Docs: best combination of speed and intelligence
AccessClaude apps (claude.ai); Claude Platform API as claude-sonnet-5-5; AWS, Google Cloud and Microsoft Azure.
Price$2 input / $10 output per 1M tokens
22 Sep 2026Frontier LLM
Claude Opus 5.5
The first model in Anthropic's Claude 5.5 family. Anthropic says it performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5, and its docs recommend it as the starting model for most workloads. External evaluators including METR tested it before release.
- 1M-token context window, 128K max output (docs)
- Terminal-Bench 4.0: 66.4% (Anthropic-reported)
- Fast mode at $8 / $40 per 1M tokens
- Knowledge cutoff June 2026 (docs)
AccessClaude apps, Claude Code and Claude Cowork; Claude Platform API as claude-opus-5-5; AWS, Google Cloud and Microsoft Azure.
Price$4 input / $20 output per 1M tokens
1 Sep 2026Frontier LLM
Claude Fable 5.1
Anthropic's most advanced generally available model, released together with Claude Mythos 5.1, which Anthropic says is the same model with different safeguard settings. Anthropic positions the pair as its most advanced models for coding and knowledge work. Fable 5.1 cut typical workload costs by about 25% compared with Fable 5.
- 1M-token context window, 128K max output (docs)
- Cache reads $0.25 per 1M tokens, 75% below Fable 5
- Mythos 5.1 variant only through trusted-access programs
- Knowledge cutoff June 2026 (docs)
AccessFable 5.1: generally available on the Claude API, AWS, Google Cloud and Microsoft Azure. Its twin Claude Mythos 5.1 is restricted to the Cyber Verification Program and the Life Sciences Verification Program.
Price$10 input / $50 output per 1M tokens
24 Jul 2026Frontier LLM
Claude Opus 5
The Opus model of Anthropic's Claude 5 generation, which Anthropic said comes close to Claude Fable 5's intelligence at half the price. Anthropic reported it as state of the art on its coding and knowledge-work evaluations while trailing Mythos 5 on cybersecurity. It became the default model on Claude Max.
- Fast mode about 2.5x quicker at double the price
- Same list price as Claude Opus 4.8
- Default model on Claude Max at launch
AccessClaude apps (default model on Claude Max, top option for Claude Pro), Claude API and cloud platforms.
Price$5 input / $25 output per 1M tokens
30 Jun 2026Mid-size LLM
Claude Sonnet 5
The Sonnet model of Anthropic's Claude 5 generation, which Anthropic calls its most agentic Sonnet yet. It can make plans, use tools such as browsers and terminals, and run autonomously at a level that previously needed larger, more expensive models. It became the default model for Free and Pro users.
- Plans and uses tools such as browsers and terminals
- Runs long autonomous agent tasks
- Default model on Claude Free and Pro plans
AccessClaude apps (default for Free and Pro; also Max, Team and Enterprise); Claude API as claude-sonnet-5.
Price$2 input / $10 output per 1M tokens
9 Jun 2026Frontier LLM
Claude Fable 5
Anthropic's first generally available Mythos-class model, launched alongside the restricted Claude Mythos 5. Anthropic said its capabilities exceeded any model it had made generally available. Access to both models was suspended on 12 June after a US Commerce Department export-control directive and restored on 1 July 2026.
- Some cyber-related queries routed to Claude Opus 4.8 by safeguards
- Mythos 5 variant for Glasswing partners with cyber safeguards lifted
- Software engineering, knowledge work, vision and science focus
AccessClaude API and paid plans (phased rollout to Jun 22) and consumption-based Enterprise; twin Claude Mythos 5 restricted to Project Glasswing partners. Access to both was suspended Jun 12 and restored Jul 1, 2026.
Price$10 input / $50 output per 1M tokens
28 May 2026Frontier LLM
Claude Opus 4.8
An Opus update that builds on Opus 4.7 with stronger coding, agentic and professional-work performance and more consistency on long-running tasks. Anthropic said it is about four times less likely than its predecessor to let flaws in code it wrote pass unremarked. It added effort controls and dynamic workflows.
- Effort controls to trade quality against speed
- Dynamic workflows for large parallel tasks
- Fast mode at $10 / $50 per 1M tokens
Accessclaude.ai; Claude API as claude-opus-4-8; Claude Code on Enterprise, Team and Max plans.
Price$5 input / $25 output per 1M tokens
16 Apr 2026Frontier LLM
Claude Opus 4.7
A generally available Opus update that Anthropic described as a notable improvement on Opus 4.6 in advanced software engineering, especially on the hardest tasks. Anthropic said it handles long-running work with rigor, follows instructions precisely and devises ways to verify its own outputs. It also improved vision.
- Reads images up to 2,576 pixels on the long edge
- Long-running coding tasks with self-verification
- Available on three major cloud platforms
Accessclaude.ai; Claude API as claude-opus-4-7; Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry.
Price$5 input / $25 output per 1M tokens
7 Apr 2026Frontier LLM
Claude Mythos Preview
An unreleased general-purpose frontier model that Anthropic shared only with Project Glasswing partners for defensive security work. Anthropic said it can surpass all but the most skilled humans at finding and exploiting software vulnerabilities, and that it had already found thousands of high-severity flaws, including in every major operating system and browser.
- CyberGym vulnerability reproduction 83.1% vs Opus 4.6 66.6% (Anthropic)
- Found long-standing bugs in OpenBSD, FFmpeg and the Linux kernel
- Backed by $100M in usage credits for participants
AccessRestricted access: Project Glasswing only (12 launch partners plus 40+ organizations that maintain critical software), via Claude API, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry. Not generally available.
Price$25 input / $125 output per 1M tokens (Glasswing participants)
15 Oct 2025Small/fast LLM
Claude Haiku 4.5
Anthropic's small, fast model, which it said delivers coding performance similar to Claude Sonnet 4 at one-third of the cost and more than twice the speed. It is still the current Haiku model in Anthropic's docs; the Sonnet 5.5 announcement says a Claude Haiku 5.5 is coming soon.
- 200K-token context window, 64K max output (docs)
- SWE-bench Verified 73.3% (Anthropic-reported)
- Docs: fastest model with near-frontier intelligence
AccessClaude apps and Claude Code; Claude API as claude-haiku-4-5-20251001; Amazon Bedrock and Google Cloud Vertex AI.
Price$1 input / $5 output per 1M tokens