GitHub made Project HydraFusion available for AI model orchestration in the experimental mode of Copilot CLI
Since 7 September, Project HydraFusion has been available in experimental mode (/experimental) in GitHub Copilot CLI. It is a system that automatically selects and combines AI models for a given task; according to GitHub, it achieved higher quality at lower costs than Claude Opus 5 in the TerminalBench 2.1 test.
GitHub made Project HydraFusion available from 7 September in the experimental mode of GitHub Copilot CLI, accessible through the /experimental command. Users select HydraFusion like any other model, and the system then decides how to process the task itself — without manual configuration, it routes requests between local, cloud and combined models with the aim of balancing performance, costs and latency. The feature is part of a broader strategy by GitHub for automatic semantic routing in Copilot.
According to GitHub, this is a research preview, meaning an experimental deployment rather than a production release. For each request, the system chooses one of three approaches — standalone processing by a single model, a cascade with escalation to a more capable model, or a critique in which a second model independently assesses the result.
In offline tests conducted by GitHub, HydraFusion achieved verified task quality 4.9 percentage points higher in the TerminalBench 2.1 benchmark at 67 % lower estimated costs compared with Claude Opus 5. In the DeepSWE benchmark, it was 1.5 percentage points behind Claude Opus 5 with 36% cost savings, and in the internal CheckpointBench test, it lagged by 0.1 percentage points with 65% savings. GitHub cautions that these are controlled offline results tied to specific benchmark versions, configuration and pricing assumptions, and that it has yet to verify actual behavior in production.
Why it matters
Developers using GitHub Copilot CLI can now enable automatic routing between models instead of switching manually, which, according to tests by GitHub, may reduce the costs of agentic coding tasks while delivering comparable or better quality. However, this is an experimental feature (research preview), so benchmark results may not directly reflect behavior in everyday use.
What was added since the original report
Verified updates
-
Specific rollout date: 2026-09-07 (missing from the existing event); Available in experimental mode through the /experimental command; Part of GitHub Copilot CLI (scope clarification); Can be selected like any other model in the interface; Automatic routing without manual configuration by the user
- Specific rollout date: 2026-09-07 (missing from the existing event)
- Available in experimental mode through the /experimental command
- Part of GitHub Copilot CLI (scope clarification)
- Can be selected like any other model in the interface
- Automatic routing without manual configuration by the user
Two audiences, two different impacts
What this means
For individuals
Developers using GitHub Copilot CLI can enable HydraFusion through the /experimental command and let the system automatically choose between models based on the task, instead of selecting a model manually.
For a business
Companies deploying GitHub Copilot can expect lower costs for agentic coding tasks according to tests by GitHub, but the feature is still a research preview, so it is not intended to be fully relied on in production.
DevelopmentCheck the original
Event sources
clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.