--- title: "Foxl v0.2.5: Claude Opus 4.7 Day-One Support" description: "Historical release notes for Foxl's initial Claude Opus 4.7 model registration, routing, and adaptive reasoning configuration." author: "Foxl Team" date_published: "2026-04-16" date_modified: "2026-07-16" canonical_url: "https://foxl.ai/blog/v0-2-5" markdown_url: "https://foxl.ai/blog/v0-2-5/index.md" image: "https://foxl.ai/blog/v0-2-5-model-contract.png" social_image: "https://foxl.ai/blog/social/v0-2-5.png" content_type: "Release note" topics: ["Model catalog","Amazon Bedrock","Adaptive reasoning"] products: ["Foxl Agent"] --- # Foxl v0.2.5: Claude Opus 4.7 Day-One Support > At its April 16 launch, Foxl v0.2.5 added Claude Opus 4.7 on Bedrock as the default model and expert-routing target, with adaptive thinking and Opus 4.6 retained for existing selections. - Author: Foxl Team - Published: 2026-04-16 - Last reviewed: 2026-07-16 - Reading time: 2 minutes - Canonical HTML: [https://foxl.ai/blog/v0-2-5]() ![The v0.2.5 model contract registering Claude Opus 4.7 with routing, reasoning, and fallback behavior.]() Foxl v0.2.5 adds Claude Opus 4.7 on AWS Bedrock and makes it the default model and expert routing target at launch. Opus 4.6 remains available for existing selections. > **Correction (July 16, 2026):** an earlier revision attributed benchmark scores, a tokenizer multiplier, a 2,576-pixel vision limit, multi-session memory behavior, and an `xhigh` effort level to Opus 4.7. Those claims were not verified by Foxl's release source. The quoted benchmark pair was later recorded for Opus 4.8 in v0.2.45, while Foxl's Opus controls support Low through Max. The unsupported claims have been removed. ## What changed in Foxl The concrete change is the integration into [Foxl's model catalog](): Bedrock model IDs, model-selector support, routing, usage accounting, and adaptive-thinking configuration. At launch, Auto could choose Opus 4.7 as its expert-level target. Opus 4.7 rejects the older budget-enabled thinking shape. Foxl now uses adaptive thinking for Opus and Sonnet 4.6 or newer, while Haiku 4.5 keeps its budget-based mode. > **Launch follow-up (April 16, 2026):** v0.2.6 corrected the context gauge from 200K to the catalog's 1M window and updated the Bedrock SDK for the new reasoning block. v0.2.7 then requested summarized thinking so reasoning text appeared instead of an empty signed block. > **Model-selection update (May 28 and June 10, 2026):** v0.2.45 made Opus 4.8 the default while keeping an explicit Opus 4.7 choice. v0.3.6 removed Auto because complexity-based routing was not reliable enough. Current builds use the model you select. ## Stability v0.2.5 also prevents an Electron shutdown crash when the tunnel client writes to a server pipe that has already closed. ## Known limitations The first v0.2.5 build did not display context usage or reasoning correctly until the two patch releases above. Bedrock model availability also depends on the AWS account and region in your [AWS Bedrock provider configuration](). A model appearing in the catalog does not by itself prove that the provider can serve it. ## Upgrade and verify Use at least v0.2.7 when evaluating the original Opus 4.7 integration. Refresh Providers, select Opus 4.7 explicitly, send a reasoning-enabled prompt, and confirm that reasoning text and context usage both appear. On a current build, do not assume the default still means 4.7; read the model label before sending. Read the [full Foxl changelog]() or download the [Foxl v0.2.5 GitHub release](). ## References and further reading 1. [Current model catalog]() - Documentation 2. [Provider configuration]() - Documentation