#ollama local ia model rtx 4070

1 messages · Page 1 of 1 (latest)

vagrant sphinx
#

How fix this please
'''
⚠️ Agent failed before reply: Model context window too small (8192 tokens). Minimum is 16000.
Logs: openclaw logs --follow
'''

boreal galleonBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

copper adderBOT
# vagrant sphinx How fix this please ''' ⚠️ Agent failed before reply: Model context window too ...

That error means the Ollama model OpenClaw is trying to use only has an 8k context window (8192), but OpenClaw hard-requires ≥ 16k to run the agent/tool prompts reliably.

Fix (pick one)

A) Switch to an Ollama model that has ≥16k context

  1. Check what OpenClaw sees:
    openclaw models list
  2. Pick an ollama/... model that shows contextWindow >= 16000, then set it:
    openclaw config set agents.defaults.model.primary "ollama/<model-id>"
openclaw gateway restart

B) If OpenClaw keeps showing 8192 for everything (auto-discovery default)
Ollama sometimes doesn’t report context length, and OpenClaw then defaults to 8192.