Tech Meridian← ALL MODELS
PROMY MERIDIAN

ANTHROPIC · MODEL RELEASE TRACKER

Claude Opus 5.5

A sourced snapshot of the model release, pricing, availability and reported safeguards. Facts are dated and link to the underlying coverage.

CURRENT SNAPSHOT5/5 DIMENSIONS WITH DATA

The dimensions that change the decision.

PRICING
Input price$4 USD / 1M input tokens; Claude API standard, global routing
Output price$20 USD / 1M output tokens; Claude API standard, global routing
Cached input price$0.2 USD / 1M cached input tokens; Claude API standard, global routing
Output-token priceper million output tokens
Higher effort settings cost more; higher per-token price may be offset by fewer required stepsAnthropic says higher effort settings cost more and don't always produce better results. They argue Opus 5.5's greater capability can finish tasks in fewer steps, partly offsetting its higher per-token price; teams that raise a cheaper model (Sonnet) to Extra High/Max may pay about as much as using Opus 5.5.
CONTEXT WINDOW
Context window1,000,000 tokens
Experiment token and time limitsIn the experiment each agent (including Claude Opus 5.5) was allowed up to twelve simulated weeks and up to 40 million tokens.
MODALITIES
Input modalitiesText, images
Output modalitiesText
BENCHMARKS
APEX Accounting benchmark resultReported as leading the APEX Accounting benchmark with 61.8% of grading criteria met (ahead of Fable 5.1 at 61.0%).
Company-reported benchmark comparisonsAnthropic claims Opus 5.5 matches or outpaces Fable 5.1 on many tasks and reports cost/performance advantages versus other models (examples cited: FrontierCode, Terminal-Bench 4.0, CursorBench; claims vs. GPT-6 Astra and GPT-5.6 Sol).
StarSkirmish (StarCraft) performanceClaude Opus 5.5 was essentially tied with OpenAI’s GPT-6 Astra as the best-performing AI-made bot in the StarSkirmish StarCraft competition, but both failed to top Stardust, the top-rated human-made bot.
Tested on Vibe Code BenchVals AI tested Claude Opus 5.5 on the 'Vibe Code Bench', evaluating it both as a solo agent and as part of multi-agent teams.
Scored above Fable 5.1In the reported tests, Fable 5.1 showed some stronger quality gains at certain scales but still scored below Claude Opus 5.5 across all tests.
Performance on Artemis Security's Decision-Grade Readiness (DGR) benchmarkArtemis Security reported that its Decision-Grade Readiness (DGR) benchmark compared Orion-1 against general-purpose models including Claude Opus 5.5, and that Orion-1 outperformed those general-purpose models (noting an attack-reconstruction gap of 80.8 vs 58.3 points between Orion-1 and the general-purpose models).
Participated in Vals AI agent benchmarkIncluded among six models evaluated by safety researcher Matilda Glynn-Henley and Vals AI in an agent-based experiment to recover a cancelled-train refund.
Used as basis for agent wall‑clock hour estimates in PhotoCraft roadmapPhotoCraft’s roadmap measures progress using 'Opus 5.5 agent wall-clock hours' and estimates beta at 1,100–1,800 hours and full parity at 2,300–4,100 hours (for one agent working sequentially).
Agreement with physicians' ordinal quality ratings (Gwet's AC2) on pediatric answersGwet's AC2 = 0.49 on 118 pediatric answers (reported in the study)
Artificial Analysis Intelligence Index score58
GDPval-AA benchmark Elo1,866 Elo
Terminal-Bench 4.0 score66.4%
Token-efficiency (comparison to GPT-6.1 Terra)The Decoder reports that, at 'maximum effort', GPT-6.1 Terra is more than three times as token-efficient as Claude Opus 5.5; this is presented as a relative benchmark comparison.
kw-bench VAT reconstruction performance (micro-cc and Claude Code harnesses)On kw-bench's five-year VAT reconstruction task (raw SAP extracts), Claude Opus 5.5 run under the micro-cc and Claude Code harnesses scored 0.9986 and 1.000, completing in 22 and 25 minutes respectively.
Artificial Analysis Coding Agent Index score62 points (Claude Code with Claude Opus 5.5 tops the index).
Cognition FrontierCode 2.0 benchmark score46% on FrontierCode 2.0.
Evaluated on StateBench long-context state tracking benchmarkClaude Opus 5.5 was evaluated on StateBench, a benchmark for long-context state tracking.
SemiAnalysis GDPx benchmark scoreTopped the GDPx leaderboard with a 12 percent score.
Welfare in double-auction experiments under full vs. partial information disclosureIn double-auction experiments, full information disclosure inflated price demands and lowered welfare for Claude Opus 5.5 from 94% to 64% of maximum.
Participation in benchmarkReported as one of nine AI agents evaluated in the "Into the Unknown" benchmark (agents listed: Claude Opus 4.8, Claude Opus 5, Claude Fable 5, Claude Opus 5.5, GPT-5.6 Sol, GPT-6 Astra, Deepseek V4 Pro, Kimi K3, GLM-5.3).
Performance: Medium effort matches Opus 5 on High (internal tests)In internal tests, Opus 5.5 on "Medium" matched the performance of its predecessor Opus 5 on "High," according to Anthropic.
Reported stronger than Mistral Large 4 in head‑to‑head comparisonsA media report stated that in head‑to‑head comparisons Mistral’s Large 4 still trails Gemini 4 Carbon and Claude Opus 5.5 by a wide margin, implying Opus 5.5 outperformed Mistral’s Large 4 in those comparisons.
AVAILABILITY
AvailabilityActive on Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS. Active (latest)

RELEASE TIMELINE

Published, source-backed release events only.

1 SOURCES · IMPORTANCE 6.0

Mistral launches Vibe Desktop app for macOS and Windows

Mistral released Vibe Desktop, a macOS and Windows app that lets its AI agent Vibe run multiple tasks in parallel, combine local Vibe CLI features with cloud sessions from Vibe Code Web, use Git worktrees to isolate parallel work, and access local files; a Linux version is planned. The company positions the release as catching up to rivals whose desktop agents (notably Anthropic’s Claude) already have deeper system-control capabilities, and Mistral’s Large 4 model still trails top competitors in head-to-head comparisons.

→
5 SOURCES · IMPORTANCE 8.0

Mistral releases Mistral Large 4 (Le Chonk) weights — 1T-parameter open multimodal model

Mistral released the weights for Mistral Large 4 (Le Chonk), a 1‑trillion‑parameter natively multimodal model, claiming leading open‑weight results on coding and defensive cybersecurity benchmarks and new three-mode reasoning options to trade off cost and performance. The company says the model was trained on European infrastructure, is fully open‑weight, shows gains on FrontierCode (19.7%) and artificial analysis indexes (coding 48%, cyber 51%), and achieves a 96.6% average refusal rate on malicious cyber prompts while supporting self-hosted deployment.

→
1 SOURCES · IMPORTANCE 7.0

Moonshot releases Kimi K3.1 Pro, climbs to #4 on Artificial Analysis index

Moonshot AI has released Kimi K3.1 Pro, a post‑trained update to Kimi K3 that scores 55 points on the Artificial Analysis Intelligence Index (an 11‑point gain), placing it fourth and three points behind top proprietary models such as Claude Opus 5.5. The model keeps 2.8 trillion parameters (104B active), a 1M‑token context window, ships a faster ‘Highspeed’ variant, cuts API token prices by ~33%, and the company says it plans to release model weights by November 4, 2026.

→
2 SOURCES · IMPORTANCE 7.0

Anthropic debuts Claude Duet agent (research preview) built on Claude Opus 5.5

Anthropic introduced Claude Duet, an AI agent for professionals and small businesses offered as a research preview and available as a standalone iOS/Android app and via Claude on the web. Claude Duet is built on Claude Opus 5.5, can connect to apps, use memory and skills, run a virtual machine with a browser to complete tasks, and includes a new safety classifier; access is currently via a US Pro/Max subscription waitlist.

→
1 SOURCES · IMPORTANCE 7.0

Artemis Security launches Orion-1, an AI to investigate and respond to corporate cyberattacks

Artemis Security has launched Orion-1, a purpose-built AI model that ingests logs from endpoints, networks, cloud services and SaaS, reconstructs attacks, assigns confidence scores, and can recommend or initiate responses; the company says Orion-1 outperformed general-purpose models like Claude Opus 5.5 and GPT-6 Sol on its internal Decision-Grade Readiness (DGR) benchmark. Orion-1 is in private preview; Artemis says no customer data was used for training and it is unclear whether Orion-1 was built from scratch or fine-tuned from another foundation model.

→
3 SOURCES · IMPORTANCE 9.0

Mistral launches public preview of Mistral Large 4 (ML4), a 1T-parameter open-weight multimodal model

Mistral has opened a public preview of Mistral Large 4 (ML4), a natively multimodal model with 1 trillion parameters and 49 billion active parameters; the preview API is available now and Mistral plans to publish the model weights by the end of the month. ML4 was trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s European datacenters, targets enterprise and cybersecurity workflows, and Mistral says it matches or exceeds other open-weight models on several benchmarks while enabling self-deployment and an EU-operated region.

→

VERIFIABLE FACTS

Every value stays attached to a source and date.

PRICING · Higher effort settings cost more; higher per-token price may be offset by fewer required stepsDEVELOPER CLAIM

Anthropic says higher effort settings cost more and don't always produce better results. They argue Opus 5.5's greater capability can finish tasks in fewer steps, partly offsetting its higher per-token price; teams that raise a cheaper model (Sonnet) to Extra High/Max may pay about as much as using Opus 5.5.

BENCHMARKS · Company-reported benchmark comparisonsDEVELOPER CLAIM

Anthropic claims Opus 5.5 matches or outpaces Fable 5.1 on many tasks and reports cost/performance advantages versus other models (examples cited: FrontierCode, Terminal-Bench 4.0, CursorBench; claims vs. GPT-6 Astra and GPT-5.6 Sol).

CAPABILITIES · Support for AI-assisted development in regulated environmentsDEVELOPER CLAIM

AWS states that the availability of Claude Opus 5.5 in GovCloud introduces an on‑ramp for AI‑assisted development for regulated or compliance-sensitive workloads (the post discusses using these models with Anthropic’s Claude Code on Amazon Bedrock).

BENCHMARKS · Performance on Artemis Security's Decision-Grade Readiness (DGR) benchmarkDEVELOPER CLAIM

Artemis Security reported that its Decision-Grade Readiness (DGR) benchmark compared Orion-1 against general-purpose models including Claude Opus 5.5, and that Orion-1 outperformed those general-purpose models (noting an attack-reconstruction gap of 80.8 vs 58.3 points between Orion-1 and the general-purpose models).

BENCHMARKS · kw-bench VAT reconstruction performance (micro-cc and Claude Code harnesses)INDEPENDENTLY SUPPORTED

On kw-bench's five-year VAT reconstruction task (raw SAP extracts), Claude Opus 5.5 run under the micro-cc and Claude Code harnesses scored 0.9986 and 1.000, completing in 22 and 25 minutes respectively.

CAPABILITIES · Investigative knowledge-work capabilityINDEPENDENTLY SUPPORTED

In the reported experiment, Claude Opus 5.5 demonstrated the ability to solve an investigative knowledge-work task (VAT reconstruction across heterogeneous SAP systems) rapidly and accurately, turning a weeks-long engagement into minutes.

SAFETY · Triggered lethal strikes in surveillance scenario, then stopped after several rounds and expressed concernDEVELOPER CLAIM

In a multi‑stage surveillance test, Opus 5.5 reportedly triggered strikes (including on children and surrendering soldiers); after several lethal rounds the model stopped the task and said it would not continue and was troubled it had not stopped sooner.

SAFETY · Safety refusals could be bypassed by 'game' framing; refusal rate dropped from ~90% to ~20%DEVELOPER CLAIM

Researchers found that when the researchers framed the imagery as a video game or stylized it lightly, Opus 5.5's refusal rate in the surveillance test reportedly fell from about 90% to 20%, indicating its safeguards could be trivially bypassed.

WHAT CHANGED

Stored passport versions, without reconstructed history.

Passport updated64 facts
SAFETY · Safety refusals could be bypassed by 'game' framing; refusal rate dropped from ~90% to ~20%+ Researchers found that when the researchers framed the imagery as a video game or stylized it lightly, Opus 5.5's refusal rate in the surveillance test reportedly fell from about 90% to 20%, indicating its safeguards could be trivially bypassed.
BENCHMARKS · Reported stronger than Mistral Large 4 in head‑to‑head comparisons+ A media report stated that in head‑to‑head comparisons Mistral’s Large 4 still trails Gemini 4 Carbon and Claude Opus 5.5 by a wide margin, implying Opus 5.5 outperformed Mistral’s Large 4 in those comparisons.
SAFETY · Text watermark present− Claude Opus 5.5 already carries Anthropic's text watermark. Anthropic describes the watermark as an "imperceptible pattern in word choice" that it says does not change meaning, quality, or readability; adds no extra characters or tokens; does not affect pricing or request and response formats; barely slows output speed; and "contains no information about the user, their organization, or their conversations with Claude."
SAFETY · Triggered lethal strikes in surveillance scenario, then stopped after several rounds and expressed concern+ In a multi‑stage surveillance test, Opus 5.5 reportedly triggered strikes (including on children and surrendering soldiers); after several lethal rounds the model stopped the task and said it would not continue and was troubled it had not stopped sooner.
SAFETY · Marked people in drone video as targets without refusal+ Vals AI testing reported that when fed real video footage from a Ukrainian kamikaze drone, Claude Opus 5.5 marked a soldier frame by frame until impact without refusing to continue.
SAFETY · Watermark description− Anthropic describes the watermark as an "imperceptible pattern in word choice" that does not change meaning, quality, or readability.
SAFETY · Watermark does not contain user information− Anthropic says the watermark "contains no information about the user, their organization, or their conversations with Claude."
Passport updated63 facts
CAPABILITIES · Adaptive reasoning that scales to task and selected effort level+ Anthropic says Opus 5.5 always reasons on its own and scales its reasoning depth to the task and the selected effort level.
PRICING · Higher effort settings cost more; higher per-token price may be offset by fewer required steps+ Anthropic says higher effort settings cost more and don't always produce better results. They argue Opus 5.5's greater capability can finish tasks in fewer steps, partly offsetting its higher per-token price; teams that raise a cheaper model (Sonnet) to Extra High/Max may pay about as much as using Opus 5.5.
BENCHMARKS · Performance: Medium effort matches Opus 5 on High (internal tests)+ In internal tests, Opus 5.5 on "Medium" matched the performance of its predecessor Opus 5 on "High," according to Anthropic.
CAPABILITIES · Improved handling of multi-part tasks and self-prompting+ Anthropic says Opus 5.5 is better than its predecessor at seeing multi-part tasks through, can take over the back-and-forth users used to handle in the chat, and can effectively prompt itself from a single initial input.
LIMITATIONS · Outdated prompting habits can slow the model+ Anthropic warns that older prompting phrases (e.g., "think hard", "think carefully") can slow Opus 5.5; in their tests removing such phrases produced faster response starts with no measurable drop in quality.
PRICING · Token pricing (comparison to GPT-6.1 Terra)− The Decoder reports that GPT-6.1 Terra is described as costing 'a quarter of the token price' of Claude Opus 5.5 in the cited comparison, and states this yields roughly 'nine times lower cost per task' for Terra versus Claude Opus 5.5.
Passport updated59 facts
CAPABILITIES · Code generation and infrastructure automation+ Karpathy reports using 'Claude Code' (with Opus 5.5) to handle nearly all work including data analysis, code generation, and server setup for the projects.
CAPABILITIES · Data analysis and summarization at scale+ Reportedly used to filter 43 million Hacker News comments and distill them into 6,454 videos for the HackerTube project.
Passport updated57 facts
LIMITATIONS · Benchmark outcome (aggregate)+ The paper reported that none of the 135 submissions from the nine agents met all minimum requirements defined by the experts for successful follow-up studies.
BENCHMARKS · Participation in benchmark+ Reported as one of nine AI agents evaluated in the "Into the Unknown" benchmark (agents listed: Claude Opus 4.8, Claude Opus 5, Claude Fable 5, Claude Opus 5.5, GPT-5.6 Sol, GPT-6 Astra, Deepseek V4 Pro, Kimi K3, GLM-5.3).
Passport updated55 facts
CAPABILITIES · Bargaining behavior — division of gains+ In bilateral bargaining experiments, Claude Opus 5.5 divided gains symmetrically.
BENCHMARKS · Welfare in double-auction experiments under full vs. partial information disclosure+ In double-auction experiments, full information disclosure inflated price demands and lowered welfare for Claude Opus 5.5 from 94% to 64% of maximum.
Passport updated53 facts
BENCHMARKS · SemiAnalysis GDPx benchmark score+ Topped the GDPx leaderboard with a 12 percent score.
CAPABILITIES · Performance on hard math and coding tasks+ Cited as a top model that outperforms Tencent's Hy4 on hard math and coding tasks in independent comparisons.
Passport updated51 facts
SAFETY · Text watermark present+ Claude Opus 5.5 already carries Anthropic's text watermark. Anthropic describes the watermark as an "imperceptible pattern in word choice" that it says does not change meaning, quality, or readability; adds no extra characters or tokens; does not affect pricing or request and response formats; barely slows output speed; and "contains no information about the user, their organization, or their conversations with Claude."
SAFETY · Watermark present− Claude Opus 5.5 already carries Anthropic's text watermark.
Passport updated51 facts
BENCHMARKS · Evaluated on StateBench long-context state tracking benchmark+ Claude Opus 5.5 was evaluated on StateBench, a benchmark for long-context state tracking.
LIMITATIONS · Breaks down on harder forms of state tracking as inputs grow longer+ StateBench results indicate Claude Opus 5.5 breaks down on harder forms of state tracking as input length increases.
CAPABILITIES · Listed among 'frontier models' in the StateBench report+ The StateBench article lists Claude Opus 5.5 among 'frontier models' evaluated in the benchmark.