Anthropic has launched Claude Sonnet 5, the next generation of its Sonnet model family, and Coaley Peak is making it available inside the Owlpen platform through Anthropic API routing.
Anthropic says Sonnet 5 is available across all Claude plans, is the default for Free and Pro users, and is available to Max, Team, and Enterprise users, as well as Claude Code and the Claude Platform.
Claude Sonnet 5 is available within Coaley Peak's Owlpen platform through Anthropic's API. In Owlpen, it should be understood as a governed model route for configured workflows, not as an unrestricted general chat surface.
What Anthropic announced
Anthropic announced Claude Sonnet 5 on 30 June 2026. It describes the model as its most agentic Sonnet model yet, with improvements over Claude Sonnet 4.6 in reasoning, tool use, coding, and knowledge work.
A stronger Sonnet tier
Anthropic says Sonnet 5 narrows the gap with Opus-class models while keeping Sonnet pricing. For business teams, the practical point is that tasks previously reserved for a more expensive foundation model route may now be worth testing against Sonnet 5 first.
Agentic work is the centre of the release
Anthropic positions the release around planning, browser and terminal tool use, software engineering, and multi-step task execution. Those are the same categories where businesses need predictable agentic workflows, clear logs, review points, and fallback handling.
Available in Owlpen
Claude Sonnet 5 is available in the Coaley Peak platform via Anthropic API routing. Client workflows should still be configured with scope limits, cost controls, fallback routes, logs, and human review where the task requires it.
API changes and migration points
Anthropic's platform documentation lists the API model ID as claude-sonnet-5. It presents Sonnet 5 as a drop-in upgrade for Sonnet 4.6, but with specific behaviour changes that matter for production integrations.
1M context and 128k output
The docs say Claude Sonnet 5 supports a 1M context window by default, 128k maximum output tokens, adaptive thinking, and the same set of tools and platform features as Claude Sonnet 4.6, except Priority Tier.
Adaptive thinking and parameter changes
On Sonnet 5, adaptive reasoning is on by default. Anthropic says manual extended thinking now returns a 400 error and that non-default sampling parameters such as temperature, top_p, and top_k also return a 400 error. Existing integrations should remove those assumptions before migration.
Token counts change
Anthropic says Sonnet 5 uses a new tokenizer. Its docs state that the same input text produces approximately 30 percent more tokens than Sonnet 4.6. That does not necessarily make a workflow more expensive on its own, but it means budgets, rate limits, and truncation settings need fresh measurement.
Pricing and deployment options
Anthropic's launch post says Claude Sonnet 5 launches with introductory API pricing of $2 per million input tokens and $10 per million output tokens through 31 August 2026. From 1 September 2026, standard pricing is listed as $3 per million input tokens and $15 per million output tokens.
Model choice should follow workflow value
Lower cost than Opus-class routes does not remove the need for testing. Teams should compare Sonnet 5 against existing Owlpen routes on actual tasks, measuring output quality, review time, latency, refusal behaviour, and total end-to-end cost rather than only the vendor benchmark position.
Cloud availability broadens deployment choices
AWS says Claude Sonnet 5 is available on Amazon Bedrock and Claude Platform on AWS, with regional availability documented by AWS. For clients with specific hosting or procurement requirements, this gives another route to evaluate alongside Anthropic's native Claude Platform and Coaley Peak's configured Owlpen integration.
Safeguards and governance
Anthropic says its pre-deployment safety evaluations found Sonnet 5 improved overall on Sonnet 4.6, including agentic safety, refusals for malicious requests, and resistance to prompt injection. It also says Sonnet 5 has cyber safeguards enabled by default.
Refusals are operational behaviour
Anthropic's docs say cyber-related refusals can return as successful HTTP 200 responses with stop_reason: "refusal". Owlpen workflows that call Sonnet 5 need to treat refusals as normal product behaviour, not only as errors.
Use higher capability deliberately
A stronger model can widen the range of tasks that can be automated, but it also increases the need for responsible AI controls. That includes scoping which systems the workflow can access, recording evidence for important decisions, and deciding where a human must approve or reject an output.
Owlpen and Claude Sonnet 5
Claude Sonnet 5 is available within the Coaley Peak platform through Anthropic API routing in Owlpen. The intended use is governed workflow delivery: model routing behind defined business processes, with cost tracking, audit logs, and escalation paths.
In practice, Sonnet 5 is a strong candidate for workflows involving document review, technical analysis, software support, operational research, and multi-step tool use. It should not automatically replace every model route. Simple classification, routine extraction, and high-volume low-risk work may still be better served by cheaper model-serving paths.
Owlpen availability
Claude Sonnet 5 is available in Owlpen through Anthropic API routing. Coaley Peak will still assess each client workflow against data sensitivity, business risk, expected volume, output review needs, and the agreed delivery model.
Sources used
This article is based on Anthropic's Claude Sonnet 5 launch post, the Claude Platform Sonnet 5 documentation, Claude Platform pricing, API release notes, the Claude Sonnet 5 System Card, and AWS's Claude Sonnet 5 on AWS announcement. Dates, pricing, availability, and safety statements should be read as current to 2 July 2026.
If you would like to discuss governed AI adoption or the Owlpen platform, contact us at enquiries@coaleypeak.co.uk.
Disclaimer. This article is published by Coaley Peak Ltd for general informational purposes only. The views expressed are those of the author, Stephen Grindley, and do not constitute legal, regulatory, financial, or technical advice. Nothing in this article should be relied upon when making procurement, investment, compliance, or technology decisions. References to third-party products, platforms, and companies are for informational purposes only and do not constitute endorsement. Availability, pricing, safety, benchmark, migration, and technical claims cited are those reported by Anthropic and AWS and have not been independently verified by Coaley Peak. References to Claude Sonnet 5 and Owlpen describe Coaley Peak platform availability only and do not imply unrestricted client use without configuration, review, and applicable commercial agreement. Readers should seek independent professional advice appropriate to their specific circumstances. Information was accurate to the best of the author's knowledge at the date of publication. Coaley Peak Ltd and Stephen Grindley accept no liability for any loss or damage arising from reliance on the contents of this article.