Claude Haiku 5.5 is now a Copilot option
On October 7, 2026, GitHub made Anthropic's Claude Haiku 5.5 generally available in GitHub Copilot. The Claude Haiku 5.5 GitHub Copilot launch puts Anthropic's cheapest and fastest small model right inside the Copilot model picker, so you can switch to it without leaving your editor, terminal, or the Copilot app. If you sell your coding time, that matters: cheaper per-token models mean cheaper agent runs and healthier margins on fixed-price projects.
GitHub positions Haiku 5.5 for fast, high-volume work — subagents, quick edits, and terminal tasks. In GitHub's early testing, Haiku 5.5 matched Claude Sonnet 5 on many coding tasks while using significantly fewer tokens and steps. That is the whole pitch in one sentence: close to a mid-tier model on everyday jobs, at a fraction of the running cost.
Who gets it and where to find it
Haiku 5.5 is available to Copilot Pro, Pro+, Max, Business, and Enterprise users, and you can select it in the model picker across the full Copilot surface: Visual Studio Code, Visual Studio, the Copilot CLI, the Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse.
Two caveats: the rollout is gradual, so check back if the model is missing from your picker; and on Business or Enterprise plans, admins can gate access through the model policy in Copilot settings. Under the default setting, new models are automatically enabled unless an admin disabled the global default or this model specifically.
Billing follows Copilot's usage-based model at provider list pricing, so the model you pick directly changes what an agent run costs you.
The price tag: what Haiku 5.5 actually costs
Anthropic calls Haiku 5.5 the cheapest, fastest, and most capable small model it has ever released, and the pricing table on its official announcement backs that up. Prices are per one million tokens:
| Prompts up to 100K tokens | Prompts over 100K tokens | |
|---|---|---|
| Input | $0.10 | $0.50 |
| Output | $0.50 | $2.50 |
Anthropic says Haiku 5.5 costs around 75% less to run on average than Haiku 4.5, and notes that prompts up to 100,000 tokens made up about 90% of requests to the previous Haiku model — so most real workloads land in the cheaper tier. The model is also Anthropic's fastest to date and the first Haiku-class model with an adjustable effort setting, so you can dial a request toward lower cost or higher intelligence instead of switching model tiers entirely.
For freelance developers, the practical math is simple. Much billable coding work is not architecture — it is linting, test fixes, boilerplate, docstrings, dependency bumps, and "explain this stack trace." Those quick, repetitive workloads are exactly what Haiku 5.5 was built for.
Meanwhile: Copilot can now drive your desktop apps
A second, separate Copilot change is worth your attention. Since October 1, 2026, computer use has been in public preview in the Copilot CLI and the GitHub Copilot app on macOS and Windows. Copilot can now interact with desktop applications on your behalf: reading accessible app content and visual context, clicking controls, entering and editing text, pressing keys, scrolling, dragging, and navigating workflows across applications.
The interesting part is the target: GitHub explicitly points at legacy and GUI-only software with no API, command-line interface, or MCP integration. Every freelancer has a client whose invoicing portal or reporting tool can only be driven by hand — computer use is a path into that layer.
Copilot asks for approval before controlling an app, you can review or reset apps you have set to always allow, and organization-managed settings can disable the feature entirely. It starts in the CLI with /computer on or in the app under Settings > Computer Use.
What this means for you
If you bill by the project, these two launches combine into a small margin upgrade:
- Cut your AI tooling costs. Move routine coding, terminal work, and subagent jobs to Haiku 5.5 in the model picker. At provider list pricing, a $0.10-per-million-token input rate changes what a heavy agent month costs.
- Productize the boring stuff. Subagents and quick edits are where freelancers lose hours to unbillable work. A cheap, fast model makes it economical to automate code reviews, test generation, and refactor passes on every ticket.
- Automate desktop chores. The monthly expense report, the copy-paste dance between a client's legacy app and a spreadsheet, the slide-deck updates — the computer-use preview can reach software no API ever will. Each automated chore is an hour you stop doing for free.
One reminder: Haiku 5.5 is generally available to paid Copilot plans, while computer use is still a public preview. Test the preview on non-critical workflows first, keep the approval prompts on, and never let it touch client financials unsupervised.
How to try it today
- Open the model picker in VS Code, the CLI, or the Copilot app and look for Claude Haiku 5.5. If it is not there yet, the rollout is still reaching accounts — check back in a few days.
- Use it for bounded tasks first: terminal commands, quick edits, and subagent jobs, and compare the results against what you normally run.
Watch GitHub's own demo of switching models in the Copilot CLI: