Claude 4.7 Migration Guide

What’s New

Claude 4.7 Opus introduces three major changes:

1. Sampling parameters removed — temperature, top_p, and top_k are no longer supported and will be ignored
2. Adaptive-only thinking — thinking.budget_tokens is no longer supported; reasoning.effort and reasoning.max_tokens are ignored (adaptive thinking is used instead)
3. New 'xhigh' effort level — A new effort level between 'high' and 'max' via verbosity / output_config.effort

Sampling Parameters Removed

Claude 4.7 Opus no longer accepts temperature, top_p, or top_k. If you pass these parameters, they will be silently ignored — your request will still succeed, but the parameters will have no effect.

1
// These parameters are ignored on Claude 4.7 Opus
2
{
3
  "model": "anthropic/claude-4.7-opus",
4
  "temperature": 0.5,
5
  "top_p": 0.9,
6
  "top_k": 50,
7
  "messages": [{ "role": "user", "content": "Hello" }]
8
}

Adaptive-Only Thinking

Claude 4.7 Opus uses adaptive thinking exclusively. The model determines how much to think based on task complexity — you cannot set a fixed token budget.

This means:

• reasoning.max_tokens is ignored (adaptive used) — budget-based thinking is not available
• reasoning.effort is ignored (adaptive used) — the model decides its own reasoning depth
• The raw Anthropic thinking.budget_tokens parameter is not supported

To control overall response effort, use verbosity instead (see below).

1
// Reasoning is always adaptive on 4.7 Opus
2
{
3
  "model": "anthropic/claude-4.7-opus",
4
  "reasoning": { "enabled": true },
5
  "messages": [{ "role": "user", "content": "Solve this problem step by step" }]
6
}

1
// reasoning.max_tokens is ignored (adaptive used)
2
{
3
  "model": "anthropic/claude-4.7-opus",
4
  "reasoning": { "enabled": true, "max_tokens": 10000 },
5
  "messages": [{ "role": "user", "content": "Hello" }]
6
}
7
// ↑ Equivalent to just { "reasoning": { "enabled": true } }

1
// reasoning.effort is ignored (adaptive used)
2
{
3
  "model": "anthropic/claude-4.7-opus",
4
  "reasoning": { "enabled": true, "effort": "low" },
5
  "messages": [{ "role": "user", "content": "Hello" }]
6
}
7
// ↑ Equivalent to just { "reasoning": { "enabled": true } }

New 'xhigh' Effort Level

A new 'xhigh' effort level is available between 'high' and 'max' via the verbosity parameter. This maps to Anthropic’s output_config.effort.

1
{
2
  "model": "anthropic/claude-4.7-opus",
3
  "verbosity": "xhigh"
4
}

The full effort scale is now: low → medium → high → xhigh → max

Verbosity Is Now the Primary Control

Since sampling parameters and reasoning budgets are removed, verbosity (which maps to output_config.effort) is the main way to influence Claude 4.7 Opus behavior:

1
// Use verbosity to control response detail and effort
2
{ "model": "anthropic/claude-4.7-opus", "verbosity": "xhigh" }

Breaking Changes