Perplexity debuts multi-model Code Mode harness
Perplexity now supports orchestrating OpenAI, Anthropic, and Google frontier models within a unified Code Mode harness to power persistent autonomous coworkers. The architecture allows developers to customize reasoning depth and model selection to handle complex workflows like reproducing bugs and opening pull requests.
Perplexity is positioning itself as a model-neutral agent harness, letting developers orchestrate frontier models interchangeably rather than locking into a single provider.
- –Model neutrality prevents vendor lock-in by letting developers route tasks across GPT, Claude, and Gemini according to specific reasoning strengths
- –Code Mode executes multi-step tool interactions within sandboxed code runtimes, eliminating intermediate token bloat in the context window
- –Persistent coworkers maintain state and project context across external platforms like Slack and Linear rather than resetting between prompts
- –Autonomous end-to-end bug reproduction and PR generation pushes Perplexity directly into competition with dedicated coding agent platforms
- –Configurable reasoning levels enable fine-tuning between execution speed and deep problem-solving effort
DISCOVERED
3h ago
2026-11-22
PUBLISHED
4h ago
2026-11-22
RELEVANCE
AUTHOR
AravSrinivas