Gemini 3 Deep Think

LLM Provider / Model · Hardest reasoning, research analysis, math olympiad and competitive programming

At a glance

Input price	$5
Output price	$30
Cache price	$1
Price tier	Premium
Context window	1M+
Max output	64K
Context tier	500K+
Speed tier	Slow/Reasoning
Latency	high
Knowledge cutoff	Jan 2026
Modality	Multimodal (vision)
Model ID	gemini-3-pro-deep-think
Provider	Google
HumanEval	96%
MMLU	93%
SWE-Bench	80%
Benchmark	Top on FrontierMath, Humanity Last Exam, ARC-AGI-2
Released	Dec 2025
Hosting	Closed/API
Capabilities	Vision, Tool use, Extended reasoning, Multi-step planning, Function calling

Vision, Tool use, Extended reasoning, Multi-step planning, Function calling

Hardest reasoning, research analysis, math olympiad and competitive programming

Trae's built-in model menu may not include Gemini 3 Deep Think — Trae primarily ships with Claude/GPT/Gemini/DeepSeek.
Kiro's spec mode is tuned for Claude models — Gemini 3 Deep Think may work but spec-driven flows are best with Claude.
OpenAI Codex (VS Code) is tied to OpenAI's GPT-5.1 Codex family — Gemini 3 Deep Think is not selectable inside this integration.
Claude Code only works with Claude models — your LLM is Gemini 3 Deep Think.
Devin uses Cognition's proprietary model internally — your selected Gemini 3 Deep Think will be ignored.
OpenAI Codex (Cloud) runs GPT-5.1 Codex internally — your selected Gemini 3 Deep Think will be ignored.
The Claude Code SDK is tuned for Claude models — Gemini 3 Deep Think may work via the model adapter but loses Claude-specific features (thinking, prompt caching, sub-agents).
Lovable uses a fixed internal model — your selected Gemini 3 Deep Think will be ignored.
Replit Agent 3 uses its own model selection internally — your selected Gemini 3 Deep Think will be ignored.

Build a full stack around Gemini 3 Deep Think — Flowpicker shows compatibility warnings before you commit.