Is Gemini 3.5 Flash better than Gemini 3.1 Pro?

Yes on most agentic and coding benchmarks. Gemini 3.5 Flash outperforms Gemini 3.1 Pro on SWE-bench Pro, Terminal-Bench 2.1, and most reasoning evaluations while being substantially faster and cheaper. It is positioned as a small-but-frontier-class model — Google's pitch is that you no longer need the Pro tier for most agentic workloads.

How does Gemini 3.5 Flash compare to GPT-5.5 and Claude Opus 4.7?

On SWE-bench Pro: Claude Opus 4.7 leads at 64.3%, GPT-5.5 sits at 58.6%, and Gemini 3.5 Flash trails at 55.1%. On Terminal-Bench 2.1 (agentic terminal coding): GPT-5.5 leads at 78.2%, Gemini 3.5 Flash follows at 76.2%, and Opus 4.7 sits at 66.1%. Flash's edge is speed (up to 4x faster output tokens per second) and price. For raw quality on long-horizon coding tasks, Opus 4.7 still wins; for fast agentic loops at scale, Flash wins.

When does Gemini 3.5 Pro launch?

Google confirmed at I/O 2026 that Gemini 3.5 Pro is in testing and will launch in June 2026. Pichai said internal evaluations are 'showing great improvements' over 3.1 Pro. Until then, 3.5 Flash is the most current Gemini model available.

Quick Answer

What Is Gemini 3.5 Flash? Google's New Default AI Model (May 2026)

Q: What is Gemini 3.5 Flash?

Gemini 3.5 Flash is Google's new frontier-class agentic and coding model, launched at Google I/O 2026 on May 19, 2026. It is now the default model for the Gemini app and AI Mode in Google Search. Google says it runs up to 4x faster than other frontier models in output tokens per second, with a 1 million-token input context window and 64K output tokens. It scored 55.1% on SWE-bench Pro and 76.2% on Terminal-Bench 2.1 in agentic coding tasks.

Published: May 20, 2026

What Is Gemini 3.5 Flash? Google’s New Default AI Model (May 2026)

Gemini 3.5 Flash launched on May 19, 2026 at Google I/O and is now the default model for the Gemini app and AI Mode in Google Search. It’s Google’s first “frontier-class small model” — claiming to beat Gemini 3.1 Pro on most agentic benchmarks while running up to 4x faster and at a fraction of the cost.

Last verified: May 20, 2026

Quick facts

Property	Value
Announced	May 19, 2026 (Google I/O 2026)
Vendor	Google DeepMind
Model family	Gemini 3.5
Input context	1,000,000 tokens
Output tokens	64,000
Speed	Up to 4x faster than other frontier models (output TPS)
SWE-bench Pro	55.1% (single-attempt)
Terminal-Bench 2.1	76.2%
Default in	Gemini app, AI Mode in Search, Antigravity 2.0
Powers	Gemini Spark, Information Agents in Search

What’s actually new

Gemini 3.5 Flash isn’t a Pro-class model and Google isn’t pretending it is. The story is different: a small, fast, cheap model that hits frontier-class numbers on the workloads that matter for agents.

Agent-first — tuned for tool use, planning, and long-horizon multi-step tasks. It’s the model that powers Antigravity 2.0 by default and Gemini Spark.
4x output speed — Google says output tokens per second is up to 4x faster than competing frontier models. For agentic loops that call the model repeatedly, this changes the unit economics.
1M-token input window — same as Gemini 1.5 Pro, but in a Flash-tier model.
Beats 3.1 Pro on most evals — this is Google’s main marketing claim: a Flash model that outperforms the previous-generation Pro on agentic and coding tasks.

Benchmarks

Benchmark	Gemini 3.5 Flash	Gemini 3.1 Pro	GPT-5.5	Claude Opus 4.7
SWE-bench Pro (single attempt)	55.1%	~50%	58.6%	64.3%
Terminal-Bench 2.1	76.2%	70.3%	78.2%	66.1%
Output tokens per second	~4x baseline	baseline	baseline	baseline
Context window (input)	1M	1M	~400K	200K

Read this honestly: Opus 4.7 is still the best coding model. GPT-5.5 leads on terminal-style agentic tasks. Gemini 3.5 Flash’s superpower is speed + price + context length — it’s the model you want in the loop when you’re running thousands of agent steps.

Where Gemini 3.5 Flash shows up

Surface	What it does
Gemini app	New default model (replaces 3.1 Pro for free + Pro tiers)
AI Mode in Google Search	Powers the new conversational search box + Information Agents
Antigravity 2.0	Default model in Google’s agent-first IDE
Gemini Spark	The reasoning engine for Google’s 24/7 personal AI agent
Gemini API / Vertex AI	Generally available for developers
Jules	Backs the free tier (3 concurrent / 15 daily tasks)

Pricing

Gemini 3.5 Flash is cheaper than 3.1 Pro and Google’s stated pitch is “frontier-class quality at Flash prices.” Exact per-token API prices are listed at ai.google.dev — for most Pro-tier users, the practical impact is higher rate limits + faster responses at the same Gemini app subscription price.

The new Google AI subscription stack as of May 20, 2026:

Plan	Price	What you get
Free	$0	Limited daily Gemini 3.5 Flash usage
Google AI Pro	$19.99/mo	Higher 3.5 Flash limits, Veo 3.1, NotebookLM Pro
Google AI Ultra	$99.99/mo	5x usage of Pro, priority Antigravity, early Gemini Spark, 20TB cloud, YouTube Premium
Google AI Ultra (top tier)	$199.99/mo	20x usage of Pro, expanded Project Genie access

The $250/mo tier was reduced to $200. The $100 tier is new.

Who should care

Agent developers — Flash’s 4x output speed and 1M-token context change agentic loop economics. If you’re hitting Opus or GPT-5.5 limits on cost or latency, Flash is the obvious comparison.

Anyone using AI Mode in Google Search — you’re already using 3.5 Flash, even if you didn’t notice. The new search box, Information Agents, and conversational refinement all run on it.

Cursor / Windsurf / Antigravity users — Flash is now the default in Antigravity and is the cheaper option in Cursor 3 and Windsurf model pickers.

Jules users — the free tier moves to 3.5 Flash, the Pro and Ultra tiers still use 3.1 Pro pending the 3.5 Pro launch in June.

What’s next

Gemini 3.5 Pro — June 2026 launch, currently in testing.
Gemini Omni Flash — rolling out alongside 3.5 Flash for multimodal video generation.
Gemini Spark — beta access for Ultra subscribers in the US next week.

TL;DR

Gemini 3.5 Flash is the new floor for Google’s models, not the ceiling. It’s a Flash-tier model that beats last-gen Pro on most agentic benchmarks, runs 4x faster, and is now the default everywhere — Gemini app, AI Mode in Search, Antigravity 2.0, Jules. The actual flagship (Gemini 3.5 Pro) drops in June. For now, Flash is the workhorse and it’s a strong one.