Skip to content
CC Switch WikiUnofficial Community Wiki

Claude Code /context Shows 200K Instead of 1M - CC Switch Fix

Quick fix

In Claude Code, run /model claude-opus-4-8[1m], or set ANTHROPIC_DEFAULT_OPUS_MODEL to claude-opus-4-8[1m] in ~/.claude/settings.json.

Symptom

Section titled “Symptom”
Error output
Claude Desktop声明了1M，在实际使用中只显示200K上下文窗口了
English translation

Claude Desktop declares 1M, but actual use shows only a 200K context window.

Cause

Section titled “Cause”

The provider does support 1M, but the “capability declaration → selectable model” bridge between CC Switch and Claude Code/Claude Desktop is broken. Although the CC Switch database already sets supports1m: true for the claude-opus-4-8 route, /v1/models returns only the bare model name instead of a separate model entry with the [1m] suffix, such as claude-opus-4-8[1m]. As a result, the Claude Code/Claude Desktop model selector does not show Opus 4.8 1M and continues to manage context using the standard 200K limit.

For third-party providers such as DeepSeek, another variant occurs: the message_start.model field in the DeepSeek streaming response returns deepseek-v4-pro, but CC Switch does not map it back to claude-opus-4-8. Claude Desktop looks up the context-window size by model name. Because deepseek-v4-pro is absent from its table, it falls back to the default 200K. The Codex case is different: Codex counts the full reasoning_content history in the Chat Completions context, so multi-turn conversations can easily exceed 1M and report that the message sent to DeepSeek exceeds the model’s maximum context length.

Fix

Section titled “Fix”

  1. In the current Claude Code session, run /model claude-opus-4-8[1m] to switch to the logical variant with the [1m] suffix. The model indicator should then show Opus 4.8 1M, and the context ring denominator should change to 1M.

  2. To make all future sessions use 1M by default, edit the Claude Code user configuration at ~/.claude/settings.json (C:\Users\<用户名>\.claude\settings.json on Windows) and change ANTHROPIC_DEFAULT_OPUS_MODEL from the bare model name to the form with the [1m] suffix.

    ~/.claude/settings.json
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-8"
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-8[1m]"

  3. Restart Claude Code/Claude Desktop, open a new session, and use /context or the context ring in the interface to confirm that the denominator shows about 1M instead of 200K.

Affected Versions

Section titled “Affected Versions”
ToolClaude Code / Claude Desktop / Codex
VersionCC Switch 3.16.1 - 3.19.2（3.16.1 之前据称正常，更新后开始出现）
PlatformsWindowsmacOS

Source Issues

Section titled “Source Issues”

This page is distilled from 5 real issues

FAQ

Section titled “FAQ”
Why does Code mode still show 200K when 1M is already selected in Claude Desktop?
The 1M checkbox on the Claude Desktop 3P configuration page only declares that the route supports 1M. It does not automatically add the logical claude-opus-4-8[1m] model entry to the Claude Code model selector. Claude Code must actually select the model with the [1m] suffix to raise the context budget to 1M; otherwise, it remains at 200K.
Why does Codex with DeepSeek exceed the maximum context length after only two turns, and what can I do?
When Codex uses the Responses API, reasoning is normally excluded from the context. After CC Switch converts requests to Chat Completions format, however, the complete reasoning_content from every turn is written back into the assistant message. For reasoning models such as DeepSeek V4 Pro, thinking can reach 30K-80K tokens per turn, so a few turns can exceed 1M. The issue provides no official fix. The community suggests using an Anthropic protocol endpoint to avoid counting reasoning repeatedly.
Can CC Switch show the 1M model automatically without changing the configuration?
CC Switch 3.16.x-3.19.x does not automatically generate model IDs with the [1m] suffix in /v1/models. This is a confirmed defect. The maintainer consolidated #4525 into #3003 for centralized tracking; follow that issue for progress.
Why does Claude Code return to 200K after a restart even after I manually select claude-opus-4-8[1m]?
A manual /model switch affects only the current session. New sessions inherit ANTHROPIC_DEFAULT_OPUS_MODEL from ~/.claude/settings.json. If it still contains the bare claude-opus-4-8 name, the limit returns to 200K. Set it to claude-opus-4-8[1m] to make the change persistent.
After enabling CC Switch model mapping, the Codex apply_patch tool disappeared. Is this related?
This is another variant in the same cluster. After model mapping generates model_catalog_json, Codex treats these models as non-official GPT models and stops providing built-in tools such as apply_patch. In #5947, the maintainer states that the root cause is in the Codex client or the provider, so CC Switch cannot fix it fundamentally. The recommendation is to upgrade to the new 4.0 release and test again.
Section titled “Related problems”
Claude Official check failed: could not extract base_url - CodexCC Switch proxying and load balancing do not support an official Claude subscription — use a third-party API key. Forcing it errors and risks a ban.4 issues mergedModel list is empty; Codex cannot select the DeepSeek modelUpgrade to CC Switch 4.0 and reapply the DeepSeek preset; model list display in the Codex desktop app also depends on its version support.5 issues mergedProvider switch not taking effect / Codex request failed (cc-switch)After switching providers such as DeepSeek in cc-switch, upgrade cc-switch and use its enable/save action as the sole way to write live configuration; do not…7 issues merged

This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.