Introducing Governor, a context-driven model router for coding agents Bito Ai
Your coding agent bill, cut in half
Technical design in hours, not days 

Gemini 4 Argon vs Claude Opus 5.5 vs Sonnet 5.5 for coding

Gemini 4 Argon vs Claude Opus 5.5 and Sonnet 5.5 coding benchmark overall scores

Table of Contents

Gemini 4 Argon is Google’s newest model, and it costs the same per token as Claude Sonnet 5.5, $2 per million input tokens and $10 output. That makes the choice for a coding agent easy to frame: which one does the job better, for less money per session and in less time? We ran Argon through the same long coding sessions, prompts and grading we use for every model, next to Claude Opus 5.5 and Sonnet 5.5.

Short answer: Gemini 4 Argon scores 76% overall on our coding benchmark, level with Claude Sonnet 5.5 at 75% and well behind Claude Opus 5.5 at 87%. Argon is stronger than Sonnet on short questions (86% against 71%) and weaker on long agentic sessions (73% against 77%). It costs a bit more than Sonnet per session and is the slowest of the three.

These are our own runs on real coding work, not numbers the vendor reported.

Gemini 4 Argon vs Claude benchmark results

Model Overall score Long agentic sessions Cost per session Time per session Short sessions Cost per question
ClaudeOpus 5.5, medium (default) 87% 87% $5.42 17.2 min 90% $0.32
NewGemini 4 Argon 76% 73% $1.87 19.3 min 86% $0.28
ClaudeSonnet 5.5, medium (default) 75% 77% $1.28 6.8 min 71% $0.11
PreviousGemini 3.1 Pro 54% 49% $1.56 5.8 min 69% $0.23

Best value in each column in bold.

Scores are percentages of the maximum. The overall score weights long agentic sessions at 75% and short sessions at 25%, because long work is where teams spend most of their time and money. A long agentic session is a multi-turn coding session on a real service: trace the code, work out why something fails, plan a fix, implement it, then audit it. Short sessions are one-shot asks, like explaining how a setting gets resolved or fixing a small bug.

50%55%60%65%70%75%80%85%90%$0$1$2$3$4$5$6Cost per long agentic session (cheaper →)Overall score (better ↑)Claude Opus 5.587% · $5.42 · 17.2 minGemini 4 Argon76% · $1.87 · 19.3 minClaude Sonnet 5.575% · $1.28 · 6.8 minGemini 3.1 Pro54% · $1.56 · 5.8 min
Overall score against cost per long agentic session. Cheaper is to the right and better is up.

Is Gemini 4 Argon better than Claude for coding?

It matches Claude Sonnet 5.5 overall and trails Claude Opus 5.5. Here’s how the turns of a long session compare:

Turn in a long session Score by model
Trace the code
92%87%95%
Analyze the failure
74%69%90%
Plan the fix
75%81%81%
Implement
63%73%85%
Audit
63%73%82%

Gemini 4 ArgonClaude Sonnet 5.5Claude Opus 5.5

Argon is ahead of Sonnet on the investigative turns and behind on the turns that write and check code.

Gemini 4 Argon vs Claude Opus 5.5

Opus wins by 11 points overall and 14 on long sessions. The gap is widest where the model has to change code: Opus scored 85% on implementation and 82% on the audit of its own change, against 63% for Argon on both. On the first turn, finding the relevant code, they’re close (92% for Argon, 95% for Opus).

Argon costs about a third of Opus per session, $1.87 against $5.42, and takes slightly longer, 19.3 minutes against 17.2. On short questions Argon gets within 4 points of Opus for slightly less money but takes about three times as long.

Gemini 4 Argon vs Claude Sonnet 5.5

Overall it’s a tie, 76% for Argon and 75% for Sonnet 5.5 at its default. The split is what differs. Sonnet is 4 points ahead on long sessions, costs less ($1.28 against $1.87) and finishes in about a third of the time (6.8 minutes against 19.3). Argon is 15 points ahead on short one-shot questions, 86% against 71%.

The two have the same list price, $2 per million input tokens and $10 per million output tokens. Argon spends more per session because it thinks longer.

How fast is Gemini 4 Argon?

It’s the slowest of the three. A long session took 19.3 minutes, against 6.8 for Sonnet 5.5 and 17.2 for Opus 5.5.

Which should you use?

For long agentic coding where cost matters, Claude Sonnet 5.5 is still the better pick: a few points higher on those sessions, cheaper per session and nearly three times faster.

For quick questions about a codebase, Argon is strong, close to Opus at a slightly lower cost, if a few minutes per answer is acceptable.

For changes where being wrong is expensive, Claude Opus 5.5 leads on every turn that involves implementing or reviewing code.

If you last tried Gemini on 3.1 Pro, Argon is a large step up: 76% against 54% overall. More in our Gemini 4 Argon vs Gemini 3.1 Pro benchmark.

How we tested

The benchmark is a set of agentic coding tasks taken from real work on our own services. Every model gets the same fixed prompts in the same coding agent (Claude Code, version pinned), and every turn is graded against an answer key that was checked against the code. The judge is Claude Opus 5.5, scoring five dimensions. The answer keys tie it to facts rather than taste, and it marks Opus down plenty. We ran 8 long sessions and 16 short ones per Gemini model, through Google’s own API, with default settings and no extra prompting. Prompt caching worked for both Gemini models, with 82-91% of input read from cache. We also ran Gemini 3.1 Pro’s short sessions in OpenAI’s Codex CLI and found no meaningful difference (71% there against 69%). We’ve found models score about the same in Claude Code, Codex and Cursor. The Claude numbers are from our earlier runs on the same tasks and answer keys. Session-to-session noise is about 3 points, so treat 1 to 2 point gaps as ties.

FAQ

Is Gemini 4 Argon better than Claude for coding?

It matches Claude Sonnet 5.5 overall (76% against 75%) and trails Claude Opus 5.5 (87%). On long agentic sessions alone, Argon scored 73%, Sonnet 5.5 77% and Opus 5.5 87%.

How much does Gemini 4 Argon cost compared with Claude?

Per long agentic session in our runs: Argon $1.87, Claude Sonnet 5.5 $1.28, Claude Opus 5.5 $5.42. Argon’s list price matches Sonnet’s, $2 and $10 per million input and output tokens.

Is Gemini 4 Argon fast?

No. A long session took 19.3 minutes, against 6.8 for Sonnet 5.5 and 17.2 for Opus 5.5.

Is Gemini 4 Argon good at short coding questions?

Yes. It scored 86% on one-shot questions, ahead of Claude Sonnet 5.5 at 71% and close to Claude Opus 5.5 at 90%.

How much better is Gemini 4 Argon than Gemini 3.1 Pro?

22 points overall (76% against 54%): 24 points on long agentic sessions and 17 on short ones.

Related benchmarks

Picture of Amar Goel

Amar Goel

Bito’s Co-founder and CEO. Dedicated to helping developers innovate to lead the future. A serial entrepreneur, Amar previously founded PubMatic, a leading infrastructure provider for the digital advertising industry, in 2006, serving as the company’s first CEO. PubMatic went public in 2020 (NASDAQ: PUBM). He holds a master’s degree in Computer Science and a bachelor’s degree in Economics from Harvard University.

Picture of Amar Goel

Amar Goel

Amar is the Co-founder and CEO of Bito. With a background in software engineering and economics, Amar is a serial entrepreneur and has founded multiple companies including the publicly traded PubMatic and Komli Media.

Written by developers for developers red heart icon

This article is brought to you by the Bito team.

Latest posts

Best cheap model for coding: Inkling 2 Small vs Claude Haiku 5.5, Gemini 3.8 Flash, DeepSeek and GPT-6.1 Terra

Inkling 2 Small vs Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 for coding

Inkling 2 Small benchmark: the strongest U.S. open-weight coding model we tested, vs DeepSeek, GLM and Kimi

GPT-6.1 Terra vs GPT-6.1 Sol vs GPT-6 Luna for coding: which OpenAI model to use

GPT-6.1 Terra benchmark: how it compares to Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 for coding

Top posts

Best cheap model for coding: Inkling 2 Small vs Claude Haiku 5.5, Gemini 3.8 Flash, DeepSeek and GPT-6.1 Terra

Inkling 2 Small vs Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 for coding

Inkling 2 Small benchmark: the strongest U.S. open-weight coding model we tested, vs DeepSeek, GLM and Kimi

GPT-6.1 Terra vs GPT-6.1 Sol vs GPT-6 Luna for coding: which OpenAI model to use

GPT-6.1 Terra benchmark: how it compares to Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 for coding

From the blog

The latest industry news, interviews, technologies, and resources.

Best cheap model for coding: Inkling 2 Small, Claude Haiku 5.5, Gemini 3.8 Flash, DeepSeek and GPT-6.1 Terra overall scores

Best cheap model for coding: Inkling 2 Small vs Claude Haiku 5.5, Gemini 3.8 Flash, DeepSeek and GPT-6.1 Terra

arrow bito ai
Inkling 2 Small vs Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 coding benchmark overall scores

Inkling 2 Small vs Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 for coding

arrow bito ai
Inkling 2 Small vs DeepSeek, GLM, Kimi and MiniMax coding benchmark overall scores

Inkling 2 Small benchmark: the strongest U.S. open-weight coding model we tested, vs DeepSeek, GLM and Kimi

arrow bito ai