Best Model for GitHub Copilot

Claude Sonnet 5.5 is the best everyday model in GitHub Copilot: it scores 61.8% on Terminal-Bench 4.0, and Copilot bills it at $2/$10 per million tokens in AI credits, the same rate as GPT-6.1 Sol (58.2%). Chat, agent mode and the CLI draw down AI credits by token (1 credit = $0.01), while code completions stay unlimited on paid plans, so the model you pick for agent mode drives the bill. Claude Haiku 5.5 bills at $0.10/$0.50, a twentieth of Sonnet 5.5, for quick questions and small edits.

Read the full Copilot review → · updated 2026-10-17

Sources: Copilot supported models · Copilot models and pricing · Terminal-Bench 4.0 leaderboard · Scale SWE-Bench Pro V2 · Anthropic on Haiku 5.5

Which models does GitHub Copilot use? (October 2026 list)

GitHub Copilot lists 32 models from Anthropic, OpenAI, Google, xAI, Moonshot AI and Microsoft as of October 14, 2026. Copilot Free and Copilot Student get Auto model selection only, so Copilot picks the model. Copilot Pro ($10 a month) can choose the Claude Haiku and Sonnet models, the smaller GPT models, Gemini Flash, Grok, Kimi K3 and MAI-Code-1.1-Flash. Pro+ ($39), Max ($100), Business and Enterprise add Claude Opus and Fable, GPT-5.5, GPT-5.6 Sol, GPT-6 Astra, GPT-6 Sol, GPT-6.1 Sol and Gemini 4 Argon.

Maker	Copilot Pro	Added on Pro+, Max, Business, Enterprise
Anthropic	Claude Haiku 4.5, Claude Haiku 5.5, Claude Sonnet 5, Claude Sonnet 5.5	Claude Opus 4.8 (fast mode in preview), Claude Opus 5, Claude Opus 5.5, Claude Fable 5, Claude Fable 5.1
OpenAI	GPT-5 mini, GPT-5.3-Codex, GPT-5.4, GPT-5.4 mini, GPT-5.6 Luna, GPT-5.6 Terra, GPT-6 Luna	GPT-5.5, GPT-5.6 Sol, GPT-6 Astra, GPT-6 Sol, GPT-6.1 Sol
Google	Gemini 3.7 Flash, Gemini 3.8 Flash	Gemini 4 Argon (added October 13)
xAI	Grok 4.5, Grok 4.6, Grok 4.7	Nothing extra
Moonshot AI	Kimi K3	Nothing extra
Microsoft	MAI-Code-1.1-Flash	Nothing extra

Retired October 2, 2026: Claude Opus 4.7, Gemini 3.5 Flash, Gemini 3.6 Flash and Kimi K2.7 Code. GitHub suggests Claude Opus 5.5, Gemini 3.8 Flash and Kimi K3 instead.
Added October 13, 2026: Gemini 4 Argon, on Pro+, Max, Business and Enterprise only, billed at Google's list price.
Claude Sonnet 4.6 was retired on September 1, 2026 but stays on annual Pro and Pro+ plans. GPT-5.4 nano runs only in the Codex extension for VS Code, on Pro+.
On Business and Enterprise, admins decide which models members see. Kimi K3 and Claude Fable 5 and 5.1 stay off until an admin turns them on.
When no other model is available, Copilot falls back to GPT-5.3-Codex, its long-term support model.

Sources: Copilot supported models · October 2 retirements · Gemini 4 Argon in Copilot · AI coding plan changes

#1Claude Sonnet 5.5Anthropic · $4.40/1M · 1M ctx

Best overall. 61.8% on Terminal-Bench 4.0 at $2/$10 per million tokens in AI credits: the strongest score per credit among Copilot's models.

Overall 96/100 · Value 59Full breakdown →

#2Claude Opus 5.5Anthropic · $8.80/1M · 1M ctx

Best for long agent runs. #1 on Terminal-Bench 4.0 at 64.9%, at $4/$20 in credits. GitHub's model guide points to GPT-6 Astra and Claude Fable 5.1 for long-horizon work; both bill at $10/$50 and score lower on that board (58.2% and 57.9%).

Overall 98/100 · Value 51Full breakdown →

#3GPT-6.1 SolOpenAI · $4.40/1M · 1.1M ctx

OpenAI pick. 58.2% on Terminal-Bench 4.0 and 96.1 on Scale's SWE-Bench Pro V2, at the same $2/$10 credit rate as Sonnet 5.5.

Overall 92/100 · Value 58Full breakdown →

#4Claude Haiku 5.5Anthropic · $0.22/1M · 1M ctx

Budget pick. $0.10/$0.50 per million tokens in credits, the same as GPT-6 Luna, and far stronger on long tasks (Anthropic reports 39.2% on Terminal-Bench 4.0; Luna scores 16.4%).

Overall 82/100 · Value 97Full breakdown →

#5Gemini 3.8 FlashGoogle · $1.65/1M · 1.0M ctx

Best value for chat. $0.75/$3.75 per million tokens in credits and 94.9 on Scale's SWE-Bench Pro V2. It is weak on long agent runs (19.1% on Terminal-Bench 4.0), so keep it for chat and focused edits.

Overall 76/100 · Value 63Full breakdown →

Use other models in Copilot

Copilot Chat in VS Code can also run models that are not on this list. Choose Manage Language Models > Add Models > Custom Endpoint and add any OpenAI-compatible API: chat and agent sessions then run on that provider and are billed there instead of spending AI Credits, while inline completions stay on your Copilot plan. On Business and Enterprise an admin must first enable the Bring Your Own Language Model Key in VS Code policy. Standard Compute is one such endpoint: DeepSeek, GLM, Qwen, Claude and GPT models on one key for a fixed monthly price from $19, and requests stop at the plan's compute budget instead of billing overage.

Set up a custom endpoint in GitHub Copilot →Why Copilot AI Credits run out →

FAQ

What model does GitHub Copilot use?

By default, Auto model selection: Copilot picks among the GPT-5.6, GPT-6 and GPT-6.1 models, Claude Haiku 4.5, Sonnet 5 and 5.5, Opus 4.8, 5 and 5.5, Gemini 3.8 Flash and MAI-Code-1.1-Flash. Copilot Free and Student get Auto only. On Pro and higher you can choose a model in the picker: Pro covers the Haiku, Sonnet, smaller GPT, Gemini Flash, Grok and Kimi K3 models, and Pro+, Max, Business and Enterprise add Claude Opus and Fable, GPT-5.5, the GPT Sol and Astra models and Gemini 4 Argon.

Which model is best in GitHub Copilot?

Claude Sonnet 5.5 for everyday chat and agent mode (61.8% on Terminal-Bench 4.0 at $2/$10 per million tokens in AI credits), Claude Opus 5.5 for the hardest tasks, and Claude Haiku 5.5 to make credits last on small requests.

How does Copilot bill different models?

Since June 1, 2026, chat, agent mode and the CLI use AI credits metered by tokens at a per-model rate (1 credit = $0.01). Code completions and next edit suggestions are not billed in credits and stay unlimited on paid plans.

Is GPT-6 Astra worth it in Copilot?

Rarely. It bills at $10/$50 per million tokens, five times GPT-6.1 Sol, and the two tie at 58.2% on Terminal-Bench 4.0. Claude Opus 5.5 scores 64.9% at $4/$20.

Should I still use Claude Sonnet 4.6 or GPT-5.4 in Copilot?

No. Claude Sonnet 4.6 was retired on September 1, 2026 and stays only on annual Pro and Pro+ plans; GPT-5.4 is still listed. Their replacements have lower list prices: Claude Sonnet 5.5 at $2/$10 per million tokens against Sonnet 4.6's $3/$15, and GPT-6.1 Sol at $2/$10 against GPT-5.4's $2.50/$15.

Every model on this page runs on Standard Compute

Standard Compute

Current frontier models Claude Fable 5.1 · GPT-6 Astra