Skip to content
TechPP

Festive sale: the best deals in one sheet. See the deals

HomeUpdate TrackerClaude Updates

Claude Update Tracker

Anthropic Claude is a leading model in programming. Its latest release is Claude Haiku 5.5, launched on October 7, 2026, Anthropic's fastest Claude model, with a 1M-token context window and prices starting at $0.10 per million input tokens.

If you have been tracking the latest Anthropic model updates, you have come to the right place. In this post, we will track Anthropic model updates and update this post every time there is a new model.

As always, you can bookmark this post for the latest updates and grab your popcorn to see all the latest model updates and changes from Anthropic.

Anthropic Claude Logo

Claude Model Releases

Claude Haiku 5.5

October 7, 2026
  • NEW
    Release:Anthropic released Claude Haiku 5.5, its fastest Claude model
  • CAPABILITY
    Context:Handles a 1M-token context window with up to 128K output tokens, up from 200K and 64K on Haiku 4.5
  • FEATURE
    Thinking:Adds adaptive thinking and adjustable effort, which Haiku 4.5 lacked, with medium as the default effort
  • BENCHMARK
    Knowledge Work:Reaches 1620 Elo on GDPval-AA v2.1, up from 735 for Haiku 4.5
  • BENCHMARK
    Computer Use:Scores 72.4% on OSWorld 2.1, up from 15.7%
  • BENCHMARK
    Coding:Scores 39.2% on Terminal-Bench 4.0 and 46.4% on FrontierCode 1.1
  • BENCHMARK
    Reasoning:Scores 57.4% on Humanity's Last Exam with tools and 45.9% without
  • PRICE
    Cost:Priced at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100K tokens, rising to $0.50 and $2.50 above that
  • NEW
    Availability:Available in the Claude API as claude-haiku-5-5

Claude Sonnet 5.5

September 28, 2026
  • NEW
    Release:Anthropic released Claude Sonnet 5.5, a significant upgrade over Claude Sonnet 5 with near Opus 5.5 level performance
  • IMPROVED
    Speed:Generates output more than 30% faster than Sonnet 5, making it the fastest Sonnet yet
  • IMPROVED
    Efficiency:Needs fewer tokens for the same work, cutting the cost per task by up to 30%
  • BENCHMARK
    Coding:Scores 70.6% on Terminal-Bench 4.0, up from 10.3% for Sonnet 5, and 55.5% on CursorBench 4.0
  • BENCHMARK
    Computer Use:Scores 80.1% on OSWorld 2.1, up from 57.0%
  • BENCHMARK
    Reasoning:Scores 64.5% on Humanity's Last Exam with tools
  • BENCHMARK
    Knowledge Work:Reaches 1844 Elo on GDPval-AA v2.1, up from 1449
  • PRICE
    Cost:Pricing holds at $2 per million input tokens and $10 per million output tokens
  • CAPABILITY
    Context:Keeps the 1M-token context window and 128K max output, with a knowledge cutoff of June 2026
  • NEW
    Availability:Available in the Claude API as claude-sonnet-5-5

Claude Opus 5.5

September 22, 2026
  • NEW
    Release:Anthropic released Claude Opus 5.5, which it says performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Opus 5
  • IMPROVED
    Speed:Generates output about 30% faster than Opus 5
  • IMPROVED
    Efficiency:Matches Opus 5's quality in roughly half the turns on agentic coding tasks
  • IMPROVED
    Code Review:Caught 72% of bugs in Anthropic's code review testing, against 56% for Opus 5 at high effort
  • BENCHMARK
    Coding:Scores 66.4% on Terminal-Bench 4.0, 54.4% on FrontierCode v1.1 and 57.8% on CursorBench 4.0
  • BENCHMARK
    Computer Use:Scores 81.8% on OSWorld 2.0
  • BENCHMARK
    Reasoning:Scores 67.7% on Humanity's Last Exam with tools
  • BENCHMARK
    Knowledge Work:Reaches 1846 Elo on GDPval-AA v2.1
  • IMPROVED
    Competition:Anthropic says it beats GPT-6 Astra on FrontierCode at about 20% of the cost, and beats GPT-5.6 Sol by 11 points on CursorBench
  • PRICE
    Cost:Priced at $4 per million input tokens and $20 per million output tokens, down from $5 and $25 on Opus 5, with cache reads at $0.20 per million
  • FEATURE
    Safety:The first Opus model to launch with Fable 5.1-class safeguards on cybersecurity, biology and distillation, including the preserved thinking safeguard
  • FEATURE
    Alignment:Anthropic calls it the strongest-performing model it has tested to date on its automated behavioural audit
  • NEW
    Availability:Available in Claude, the Claude API as claude-opus-5-5, AWS, Google Cloud and Microsoft Azure

Claude Fable 5.1 & Mythos 5.1

September 1, 2026
  • NEW
    Release:Anthropic released Claude Fable 5.1 and Claude Mythos 5.1, calling them its most advanced models for coding and knowledge work
  • FEATURE
    Safety:Both names cover the same underlying model at different safeguard levels: Fable 5.1 ships with Anthropic's production safeguards, while Mythos 5.1 stays limited to vetted cybersecurity and life-sciences organizations in the US
  • IMPROVED
    Safeguards:Roughly 60% fewer false-positive safeguard interventions on cybersecurity tasks than Fable 5
  • BENCHMARK
    Science:Scores 52.6% on Terminal-Bench-Science 0.1, up from 24.7% for Fable 5
  • BENCHMARK
    Coding:Scores 55.8% on Terminal-Bench 4.0 against 42.0% for Fable 5, with Mythos 5.1 at 60.9%
  • BENCHMARK
    Coding:Reaches 73.4% on CursorBench 3.2.0, up from 70.5%
  • BENCHMARK
    Reasoning:Scores 60.9% on Humanity's Last Exam without tools
  • PRICE
    Cost:Pricing holds at $10 per million input tokens and $50 per million output tokens, while cache reads drop 75% to $0.25 per million
  • IMPROVED
    Cost:Anthropic says typical workloads cost about 25% less than on Fable 5, and highly agentic workloads up to 45% less
  • NEW
    Availability:Fable 5.1 is generally available in Claude, Claude Code, the Claude platform, Cursor, AWS, Google Cloud, and Microsoft Azure

Claude Opus 5

July 24, 2026
  • NEW
    Release:Anthropic released Claude Opus 5 as a more thoughtful and proactive frontier model
  • IMPROVED
    Quality:Anthropic positions Opus 5 close to Claude Fable 5 capability at roughly half the price

Claude Sonnet 5

June 30, 2026
  • NEW
    Release:Claude Sonnet 5 is Anthropic's most agentic Sonnet model yet, able to plan, use tools like browsers and terminals, and run autonomously on tasks that recently needed larger, pricier models
  • IMPROVED
    Performance:Performance is close to Claude Opus 4.8, with gains over Sonnet 4.6 in reasoning, tool use, coding, and knowledge work
  • PRICE
    Cost:Introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026, then $3 input and $15 output
  • UPDATE
    Plans:Now the default model for Free and Pro plans, and available to Max, Team, and Enterprise users
  • FEATURE
    Safety:Shows an overall lower rate of undesirable behaviors than Sonnet 4.6 and is generally safer to use in agentic contexts
  • NEW
    Availability:Claude is now generally available in Microsoft Foundry on Azure, with Opus 4.8 and Haiku 4.5 in the Messages API

Claude Fable 5 & Mythos 5

June 9, 2026
  • NEW
    Release:Claude Fable 5, a Mythos-class model that sits above the Opus tier, is now Anthropic's most capable generally available model
  • NEW
    Availability:Generally available on the Claude API, Claude Platform on AWS, Amazon Bedrock, Vertex AI, and Microsoft Foundry
  • FEATURE
    Safety:Shipped as two products split by safety classifiers: Fable 5 routes flagged cyber, biology, chemistry, and distillation requests down to Claude Opus 4.8, while Mythos 5 keeps those capabilities available for vetted users
  • CAPABILITY
    Output:Supports up to 128K output tokens in a single response
  • PRICE
    Cost:Priced at $10 per million input tokens and $50 per million output tokens
  • UPDATE
    Plans:Included in Pro, Max, Team, and seat-based Enterprise plans at no extra cost through June 22, 2026; usage credits are required afterward

Claude Opus 4.8

May 28, 2026
  • NEW
    Release:Anthropic's new flagship model, arriving less than two months after Opus 4.7
  • IMPROVED
    Judgement:Sharper judgement, with the model asking the right questions and pushing back when a plan is not sound
  • IMPROVED
    Honesty:More honest about its own progress, more likely to flag uncertainty, and less likely to make unsupported claims
  • IMPROVED
    Reliability:4x less likely than Opus 4.7 to overlook flaws in code
  • IMPROVED
    Autonomy:Can work independently for longer stretches on agentic tasks
  • FEATURE
    Dynamic Workflows:Research preview in Claude Code runs hundreds of parallel subagents in one session for codebase-scale migrations, capped at 1,000 subagents
  • FEATURE
    Effort Control:New control lets users choose how much computational effort Claude applies, trading speed for depth
  • FEATURE
    API:Messages API now accepts system entries inside the messages array for mid-task instruction updates without breaking the cache
  • BENCHMARK
    Coding:Agentic coding score rose to 69.2% from 64.3% on Opus 4.7
  • BENCHMARK
    Computer Use:Online-Mind2Web reached 84%, beating GPT-5.5
  • BENCHMARK
    Reasoning:Multidisciplinary reasoning with tools rose to 57.9% from 54.7%
  • BENCHMARK
    Computer Use:Agentic computer use rose to 83.4% from 82.8%
  • BENCHMARK
    Knowledge Work:Knowledge work score climbed to 1,890 from 1,753
  • BENCHMARK
    Legal:First model to break 10% on the all-pass standard of the Legal Agent Benchmark
  • PRICE
    Cost:Pricing unchanged at $5 per million input tokens and $25 per million output tokens, and fast mode is now 3x cheaper than on previous models

Code with Claude 2026

May 6, 2026
  • UPDATE
    Models:No new flagship model was announced at the keynote; the focus was on agent infrastructure
  • NEW
    Managed Agents:Claude Managed Agents gained dreaming, multiagent orchestration, outcomes, and webhooks
  • NEW
    Claude Design:Claude Design launched as an Anthropic Labs product for creating designs, prototypes, slides, and one-pagers
  • IMPROVED
    Limits:Higher Claude Code and Opus API usage limits for developers

Claude Opus 4.7

Apr 16, 2026
  • IMPROVED
    General:Better at following instructions
  • IMPROVED
    Coding:Better at advanced software engineering
  • IMPROVED
    Vision:Better at visual and multimodal tasks
  • IMPROVED
    Reliability:Better at multi-step task reliability
  • NEW
    Memory:Better at using file-based memory across long work sessions
  • FEATURE
    Reasoning:Added xhigh effort mode for deeper reasoning control
  • CAPABILITY
    Vision:Supports images up to 2,576 pixels on the long edge, more than 3x the previous limit
  • BENCHMARK
    Coding:Improved by 13% over Opus 4.6 on Anthropic’s 93-task coding benchmark
  • BENCHMARK
    Coding:Reached 70% on CursorBench vs 58% for Opus 4.6
  • BENCHMARK
    Vision:Reached 98.5% on the XBOW visual-acuity benchmark vs 54.5% for Opus 4.6
  • BENCHMARK
    General:State-of-the-art on Finance Agent and GDPval-AA in Anthropic’s release notes

Claude Opus 4.6

Feb 2026
  • IMPROVED
    General:Better at agentic coding, tool use, search, finance, and long-context reasoning
  • IMPROVED
    Context:Better at finding and using information across huge document sets
  • IMPROVED
    Context:Better at resisting context rot in very long sessions
  • BENCHMARK
    Context:Scored 76% on MRCR v2 8-needle 1M vs 18.5% for Sonnet 4.5
  • BENCHMARK
    Legal:Harvey reported 90.2% on BigLaw Bench
  • BENCHMARK
    Reasoning:Box reported 68% vs 58% baseline in its evaluation

Claude Sonnet 4.6

Feb 17, 2026
  • IMPROVED
    General:Better at coding, computer use, planning, design, and long-context reasoning
  • FEATURE
    Context:Added 1M token context window in beta
  • BENCHMARK
    General:Matched Opus 4.6 on OfficeQA in one partner test cited by Anthropic
  • BENCHMARK
    Reasoning:Box reported a 15-point improvement over Sonnet 4.5 on document-heavy reasoning tasks
  • BENCHMARK
    General:Pace reported 94% on its insurance benchmark
  • BENCHMARK
    Coding:Anthropic noted SWE-bench gains up to 80.2% with a prompt modification

Claude Opus 4.5

Nov 24, 2025
  • IMPROVED
    General:Better than Opus 4.1 for coding, agents, and computer use
  • IMPROVED
    General:Stronger for deep research, slides, and spreadsheets
  • NEW
    Positioning:Expanded Claude’s top-end model beyond coding into broader knowledge work

Claude Haiku 4.5

Oct 15, 2025
  • NEW
    Performance:Big jump for the fast model tier
  • CAPABILITY
    Coding:Delivers coding performance similar to Sonnet 4 from five months earlier
  • CAPABILITY
    Efficiency:Runs at one-third the cost and more than twice the speed
  • BENCHMARK
    Coding:Reached 90% of Sonnet 4.5 performance in Augment’s evaluation cited by Anthropic

Claude Sonnet 4.5

Sep 29, 2025
  • IMPROVED
    Coding:Better at coding than Sonnet 4
  • IMPROVED
    Agents:Stronger on complex agents
  • IMPROVED
    Computer Use:Better at computer use
  • IMPROVED
    Reasoning:Better at reasoning and math
  • BENCHMARK
    Coding:Anthropic said one internal code-editing error rate dropped from 9% to 0% versus Sonnet 4 on that test
  • NEW
    Positioning:Positioned as Anthropic’s most aligned frontier model at launch

Claude Opus 4.1

Aug 5, 2025
  • IMPROVED
    Performance:Better than Opus 4 on coding, reasoning, and agent reliability
  • PRICING
    Cost:Keeps the same API pricing as Opus 4
  • UPDATE
    Release:Refinement update rather than a full new generation

Claude 4 family

May 22, 2025
  • IMPROVED
    Memory:Better memory during long tasks
  • IMPROVED
    Tools:Better parallel tool use
  • IMPROVED
    Tools:Better extended thinking with tools
  • CAPABILITY
    Agents:65% less likely than Sonnet 3.7 to use shortcuts or loopholes in agentic tasks

Claude Sonnet 4

May 22, 2025
  • NEW
    Release:Major upgrade over Claude 3.7 Sonnet
  • IMPROVED
    Performance:Better at coding, reasoning, and following instructions
  • BENCHMARK
    Coding:Scored 72.7% on SWE-bench in Anthropic’s launch material
  • NEW
    Positioning:More practical for day-to-day use than Opus-tier models

Claude Opus 4

May 22, 2025
  • NEW
    Release:Claude’s top model at launch
  • NEW
    Focus:Built for advanced coding and long-running agent tasks
  • BENCHMARK
    Coding:Scored 72.5% on SWE-bench
  • BENCHMARK
    Agents:Scored 43.2% on Terminal-bench
  • CAPABILITY
    Agents:Designed to work for hours on complex tasks

Claude 3.7 Sonnet

Feb 24, 2025
  • NEW
    Architecture:First hybrid reasoning Claude model
  • CAPABILITY
    Reasoning:Can answer normally or use extended thinking for harder tasks
  • FEATURE
    Control:Lets developers control how much thinking the model uses
  • IMPROVED
    Coding:Better at coding and frontend web work
  • FEATURE
    Tools:Launched with Claude Code

Computer use beta

Oct 22, 2024
  • NEW
    Feature:First public Claude feature for computer control
  • CAPABILITY
    Vision:Can look at screens, move the cursor, click, and type
  • NEW
    Agents:Built for agent tasks on supported setups
  • BENCHMARK
    Performance:Reached 14.9% on OSWorld in Anthropic’s announcement, ahead of the next-best system cited there at 7.8%

Claude 3.5 Haiku

Oct 22, 2024
  • IMPROVED
    Performance:Much stronger than earlier Haiku-class models
  • BENCHMARK
    Performance:Matched Claude 3 Opus on many evaluations according to Anthropic
  • CAPABILITY
    Efficiency:Kept the fast and lower-cost Haiku profile

Updated Claude 3.5 Sonnet

Oct 22, 2024
  • IMPROVED
    Coding:Better coding performance than the earlier 3.5 Sonnet release
  • BENCHMARK
    Coding:SWE-bench Verified improved from 33.4% to 49.0%
  • BENCHMARK
    Performance:TAU-bench retail improved from 62.6% to 69.2%
  • BENCHMARK
    Performance:TAU-bench airline improved from 36.0% to 46.0%
  • IMPROVED
    Tools:Better tool use and agent-style workflows

Claude 3.5 Sonnet

Jun 21, 2024
  • NEW
    Release:Major intelligence upgrade over Claude 3 models
  • IMPROVED
    General:Better at coding, reasoning, instruction following, and vision
  • BENCHMARK
    Performance:Outperformed Claude 3 Opus on several evaluations
  • CAPABILITY
    Efficiency:Ran at around 2x the speed of Claude 3 Opus
  • BENCHMARK
    Agents:Scored 64% on Anthropic’s internal agentic coding eval vs 38% for Claude 3 Opus
  • FEATURE
    UI:Introduced Artifacts alongside the model

Claude 3 Haiku

Mar 13, 2024
  • NEW
    Release:Fastest Claude 3 model
  • NEW
    Positioning:Lower-cost model for high-volume tasks
  • CAPABILITY
    Processing:Can process dense docs and charts very quickly
  • FEATURE
    Vision:Added strong vision support at the fast tier

Claude 3 Sonnet

Mar 4, 2024
  • NEW
    Positioning:Balanced model for speed and quality
  • CAPABILITY
    Efficiency:Around 2x faster than Claude 2 and 2.1
  • NEW
    Focus:Better for production and daily use
  • FEATURE
    Context:Ships with 200K context window

Claude 3 Opus

Mar 4, 2024
  • NEW
    Release:Flagship model in the Claude 3 lineup
  • IMPROVED
    Performance:Better at reasoning, coding, analysis, and long-context tasks
  • BENCHMARK
    Context:Reached 99%+ accuracy on Anthropic’s Needle In A Haystack long-context test
  • BENCHMARK
    Performance:Set the pace for Claude on benchmarks like MMLU, GPQA, and GSM8K at launch

Claude 2.1

Nov 21, 2023
  • FEATURE
    Context:Context window increased to 200K tokens
  • IMPROVED
    Reliability:Lower hallucination rate than Claude 2
  • FEATURE
    Prompts:Added system prompts
  • FEATURE
    Tools:Added tool use in beta
  • IMPROVED
    Focus:Better for long documents and enterprise workflows

Claude 2

Jul 11, 2023
  • IMPROVED
    Performance:Better at reasoning, coding, and math
  • CAPABILITY
    Generation:Can generate longer responses than Claude 1
  • NEW
    Access:Expanded access through Claude.ai and the API

Claude 1 100K

May 11, 2023
  • FEATURE
    Context:Context window increased to 100K tokens
  • CAPABILITY
    Processing:Can handle very long documents and large prompts in one go
  • NEW
    Use Cases:Opened up research, document analysis, and enterprise use cases

Claude 1

Mar 14, 2023
  • NEW
    Release:First Claude model release
  • NEW
    Focus:Built for general chat, writing, summarising, and Q&A
  • NEW
    Positioning:Released with a stronger focus on safety and reliability
  • NEW
    Release:Claude Instant also launched as the faster and lower-cost model

Was this article helpful?

Comments

0

Share your thoughts, questions, or feedback about this guide.

Leave a Comment

0/2000