> ## Documentation Index
> Fetch the complete documentation index at: https://phaseo.app/docs/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Phaseo Gateway base URL: https://api.phaseo.app/v1. Send `Authorization: Bearer $PHASEO_API_KEY` with every request.
> Model IDs look like `provider/model`, for example `openai/gpt-6-astra`. Do not guess IDs: list them with `GET /v1/models` or check https://phaseo.app/models. `phaseo/free` routes to an available free model and needs no wallet balance.
> For new text integrations prefer `POST /v1/responses`. `POST /v1/chat/completions` (OpenAI format) and `POST /v1/messages` (Anthropic format) suit existing code.
> The full API contract is https://phaseo.app/docs/openapi/v1/openapi.yaml. Official SDKs: `@phaseo/sdk` (TypeScript) and `phaseo` (Python), plus Go, C#, PHP, Ruby and Rust.
> Search these docs over MCP at https://phaseo.app/docs/mcp, and install Phaseo skills with `npx skills add https://phaseo.app/docs`. If a page is wrong or out of date, report it with the MCP `submit_feedback` tool.
> Manage Phaseo accounts with the `phaseo` CLI (`npm install -g @phaseo/cli`, then `phaseo login` or `phaseo login --device-code`), the `/v1` Management API or the MCP server at `https://mcp.phaseo.app/mcp?access=manage`. Do not send users to the dashboard for account tasks. A person is still needed for sign-up, approving a sign-in, paying the first top-up on the checkout link that `phaseo credits add` returns when no card is saved, confirming each saved-card top-up from the emailed link, and account security (password, MFA, passkeys). Create API keys with `phaseo keys create --name <name> --json`; never guess one. Credit top-ups need `phaseo login --scopes billing:write` first. Task map: https://phaseo.app/docs/v1/developers/manage-from-an-agent.md.

# Anthropic Claude Opus 4.8

> What to know when adopting Claude Opus 4.8, including adaptive thinking, sampling controls, and long context.

# Anthropic Claude Opus 4.8

Use this guide to adopt `anthropic/claude-opus-4.8` safely in production.

Anthropic deprecated Claude Opus 4.1 on June 5, 2026 and scheduled retirement for August 5, 2026. Their migration guidance for Opus 4.1 is cumulative: first apply the Claude Opus 4.7 request-shape changes, then review the Claude Opus 4.8 behavior changes.

## What's new

* the Phaseo model ID is `anthropic/claude-opus-4.8`
* Opus 4.7 and later reject non-default sampling params like `temperature`, `top_p`, and `top_k`
* Opus 4.7 and later reject manual extended-thinking budgets; use adaptive thinking instead
* Opus 4.8 defaults `output_config.effort` to `high`
* Opus 4.8 adds mid-conversation system messages
* Opus 4.8 raises the baseline context/output ceilings to 1M context and 128K max output on the Claude API, Claude Platform on AWS, Amazon Bedrock, and Vertex AI

## Migration quickstart

### 1. Update the model ID

Set the model ID to `anthropic/claude-opus-4.8`.

### 2. Remove non-default sampling parameters

If your Opus 4.1 requests still set any of these, remove them before rollout:

* `temperature`
* `top_p`
* `top_k`

On Opus 4.7 and later, non-default values for those fields return `400`.

### 3. Replace manual thinking budgets with adaptive thinking

If you send legacy thinking payloads such as:

* `thinking: { "type": "enabled", "budget_tokens": 32000 }`

move to:

* `thinking: { "type": "adaptive" }`
* `output_config.effort = "high"` as the baseline

### 4. Rebaseline effort and long-context expectations

Opus 4.8 defaults effort to `high`, and it supports a much larger context window than Opus 4.1 on Anthropic-operated API surfaces. Re-test:

* latency budgets
* token usage
* prompt caching behavior
* long-document and long-agent traces

If you run on Microsoft Foundry, do not assume the 1M context window there at launch; Anthropic documents a 200K context window for Foundry on Opus 4.8.

### 5. Re-check instruction updates in long conversations

Opus 4.8 supports mid-conversation system messages. If your agent loop currently rewrites the full system prompt every turn, you may be able to simplify that flow and preserve more cache hits.

## What to test

* requests that still send `temperature`, `top_p`, or `top_k`
* thinking-enabled routes that previously relied on `budget_tokens`
* long-horizon agent and tool workflows
* schema-sensitive outputs at explicit `effort` levels
* token-cost and latency deltas on your highest-volume Opus 4.1 prompt classes

## Safe rollout

1. Remove sampling params and legacy thinking budgets before switching traffic.
2. Shadow Opus 4.8 on production-like prompts and tool flows.
3. Canary with a fallback path still available during the overlap window before August 5, 2026.
4. Promote only after latency, cost, and task-completion deltas are within target.

## Sources

* [Anthropic model deprecations](https://platform.claude.com/docs/en/about-claude/model-deprecations)
* [Anthropic migration guide](https://platform.claude.com/docs/en/about-claude/models/migration-guide)
* [What's new in Claude Opus 4.8](https://platform.claude.com/docs/en/about-claude/models/whats-new-claude-4-8)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.