What are the verified coding and reasoning benchmarks for Gemini 3.5 Flash?

Gemini 3.5 Flash achieves SWE-bench Verified: 63.8% (Thinking (Low)), LiveCodeBench: 47%, AIME: 74.5%. Verified sources: Gemini 3.5 Flash Specifications (Google DeepMind).

Does Swarm isolate Gemini 3.5 Flash code edits from other agents?

Yes. Swarm provisions isolated Git worktrees for every session and subagent, preventing file race conditions and keeping your main working tree clean until verification passes.

Can Gemini 3.5 Flash be assigned to specific subagent roles?

Yes. In ~/.swarm/config.json, you can configure Gemini 3.5 Flash as Primary/Planner (DAG task decomposition), Coder (implementation in isolated worktrees), or Finder (read-only symbol search).

Gemini 3.5 Flash — Specs, Context & Capabilities

Q: What is the recommended Swarm multi-agent topology for Gemini 3.5 Flash?

Swarm recommends Gemini 3.5 Flash as Finder Subagent (Tier 3: Rapid Subagent Worker). High throughput and low cost make it ideal for repository indexing, AST search, and fast unit test runs. Best paired with: Claude 3.7 Sonnet as Lead Planner.

Q: How do I run Gemini 3.5 Flash with Swarm?

Configure Gemini 3.5 Flash via CLI using: swarm ctl model set --provider google --model google/gemini-3.5-flash, or switch models interactively inside the terminal TUI with /models.

Gemini 3.5 Flash Benchmarks & Agentic Accuracy

Verified primary research & developer preference metrics

Primary Sources Verified

SWE-bench Verified

63.8%

Thinking (Low)Google DeepMind Report ↗

LiveCodeBench

47%

ThinkingLiveCodeBench ↗

LMSYS Coding Elo

1345Elo

Coding ArenaLMSYS Arena ↗

AIME 2024 / 2025

74.5%

ThinkingGoogle DeepMind Report ↗

Pricing ($ / 1M tokens)

$0.15/ $0.6

$0.025 cached readStandard API Pricing

What This Means in Real Software Development

High throughput provides instant feedback in CLI and TUI sessions.
Solid 63.8% SWE-bench Verified score at ultra-low $0.15/M token pricing.
High reliability for symbol grep, multi-file edits, and automated unit test runs.
Full 1M token context window preserves build logs without truncation.

Swarm Multi-Agent Topology

Tier 3: Rapid Subagent WorkerRole: Finder Subagent

Best Paired With: Claude 3.7 Sonnet as Lead Planner

High throughput and low cost make it ideal for repository indexing, AST search, and fast unit test runs.

Git Worktree Isolation Advantage:

Executes parallel discovery jobs across multiple worktrees concurrently.

What Developers & Practitioners Are Saying

Real-world field observations, production strengths, and developer consensus

Field Consensus

"Early high-speed Flash checkpoint that demonstrated the viability of sub-$0.10/M tokens for agentic workflows."

Verified Real-World Strengths

Extremely low cost and fast generation speed
Good basic language and syntax translation

Production Caveats & Gaps

Lower SWE-bench score (35.0%) than modern 3.8 Flash

Primary Sources & Methodology Citations

Gemini 3.5 Flash Specifications — Google DeepMind (November 2025)

Gemini 3.5 Flash model overview

Gemini 3.5 Flash is listed in the public Swarm catalog through Google Gemini. This page summarizes the verified catalog fields used by Swarm without claiming real-time access for every provider account.

Provider: Google Gemini
Family: Gemini
Context window: 1.05M tokens
Thinking support: low · medium · high
Availability status: Deprecated
Catalog ID: google/gemini-3.5-flash

Audio input
File input
Image input
Text input
Video input
Text output
Reasoning
Batch
Deprecated

Run Gemini 3.5 Flash in Swarm

Verified CLI workflow & multi-agent architecture

Quickstart Guide →

bash — swarm run

$swarm ctl model set --provider google --model google/gemini-3.5-flash

# Or switch interactively inside the terminal TUI:

>/models

Subagent Role Matrix

Assign Gemini 3.5 Flash to specialized system roles in ~/.swarm/config.json:

Primary / Planner: Orchestrates full task DAGs and multi-stage pipelines.
Coder Subagent: Generates code and unit tests inside private Git worktrees.
Finder Subagent: Performs read-only code search, symbol discovery, and mapping.

System Agent Roles →Why Swarm Architecture →