Google Challenges Anthropic with Gemini 3.5 Flash and Antigravity

Google Takes on Anthropic with Gemini 3.5 Flash and Antigravity

Since the beginning of the year, Anthropic has set the pace in the field of AI for coding. However, Google, with its new models Gemini 3.5 Flash and Antigravity, is looking to catch up. These innovations promise speed and cost that could worry current leaders like Claude Code.

With Gemini 3.5 Flash, Google is making its move in the highly competitive AI coding sector. Cheaper, equally powerful, and up to five times faster than Claude Opus 4.7 or GPT-5.5, the model presented by DeepMind has everything to appeal to developers. Add to this a reimagined experience in Antigravity and a CLI mode, and Google is openly hunting in Anthropic's territory. Here's a review of the strengths at play.

The Real Strengths of Gemini 3.5 Flash

With the arrival of Gemini 3.5 Flash, Google promises that it is no longer "necessary to sacrifice quality for latency." The model achieves a feat in the new race of AI for coding: speed. Gemini 3.5 Flash is the fastest coding model Google has released to date. According to data from Artificial Analysis, from Google AI Studio, the model reaches a speed of 277 tokens per second, compared to 65 tokens per second for GPT-5.5 (xhigh mode) and 51 tokens per second for Claude Opus 4.7 (Max reasoning). It is thus approximately 4.3 times faster than GPT-5.5 and 5.4 times faster than Claude Opus 4.7. This is a significant difference for companies looking to produce code faster, particularly for agentic systems.

The achievement is real, especially since the model is not significantly worse than the leading models of the competition. In agentic tasks, Gemini 3.5 Flash even takes the lead with 83.6% on MCP Atlas (multi-step workflows orchestrated via the MCP protocol), ahead of 79.1% for Opus 4.7 and 75.3% for GPT-5.5. On Terminal-bench 2.1 (code in terminal environment), its 76.2% clearly surpasses Opus (66.1%) and is just below GPT-5.5 (78.2%).

Interface control confirms parity once again: on OSWorld-Verified (computer use), Flash achieves 78.4%, neck-and-neck with GPT-5.5 (78.7%) and Opus (78.0%). The only drop-off is on SWE-Bench Pro (varied agentic coding tasks). With 55.1%, Gemini 3.5 Flash lags behind the 64.3% of Opus 4.7 and the 58.6% of GPT-5.5. In other words, Flash holds its own everywhere and often wins in agentic tasks, except for solving complex coding tasks autonomously. The most demanding projects will still need to go through Claude or GPT-5.5… until Gemini 3.5 Pro?

The Real Question: the Scaffold

Why has Google never really taken the lead in AI for coding? Until now, it all came down to the scaffold. The entire software layer surrounding the model matters as much, if not more, than the raw model itself. An excellent model can display mediocre coding performance if the scaffold is not optimal. Anthropic has understood this well and today takes care of both Claude Code (the scaffold) and its raw models. The results speak for themselves: the reference for the developer community, including in France, remains Claude Code.

To try to reverse the trend, Google is offering a completely new version of its coding agent Antigravity. The entire experience has been rethought for developers based on extensive field feedback. Antigravity 2.0 is nothing like the heavy and unusable IDE previously offered by Google. The tool allows for orchestrating multiple coding agents in parallel, similar to Claude Code in Claude Desktop. But the real change comes with the announcement of Antigravity CLI. Google is betting almost everything on access to the agent via the command line, which has been widely praised by developers using Claude Code. Gemini CLI will also migrate to Antigravity CLI. The agent can perform the majority of tasks supported by Claude Code (goal, routine…). Will this be enough to cannibalize Anthropic's users? Hard to say, but to keep pace, Google will need to match the breakneck speed set by Anthropic since the beginning of the year in coding, while also offering a truly effective agent (testing to come on JDN).

Google Pricing

With $1.50 for input and $9 for output per million tokens, Gemini 3.5 Flash is by far the cheapest of the three. For input, it costs 3.3 times less than Claude Opus 4.7 and GPT-5.5, both priced at $5. For output, the gap narrows but remains significant: 2.8 times cheaper than Opus 4.7 ($25) and 3.3 times cheaper than GPT-5.5 ($30). Typical Google pricing, as always. The company controls the entire chain, from hardware to model, thanks to its in-house TPUs (Google avoids the Nvidia tax).

For developers who would use Google AI subscriptions directly, Google offers four plans in France:

• AI Plus at €7.99 per month
• AI Pro at €21.99 per month
• AI Ultra (new) at €99.99 per month
• AI Ultra x5 at €219 per month

The prices align with the pricing policies of OpenAI and Anthropic. Even though Google, like its competitors, does not disclose the number of tokens included, the API pricing strongly suggests that the same level of plan (e.g., €21.99) allows for significantly more generous usage at Google.

Should You Switch to Gemini 3.5 Flash?

As with every new model or major update released in the AI coding ecosystem, the question of migration arises. How can you tell if Gemini 3.5 Flash is a good alternative for your business? Start by experimenting with a small group of developers over several weeks. If the results are as good as with Claude Code for a better price, consider a migration.

The key is to remain as agile and agnostic as possible regarding the technologies used. The verdict is clear for teams whose coding agents operate in volume on agentic workflows: the economic calculation clearly favors Google. For others, those relying on complex coding tasks autonomously, it’s better to wait for Antigravity CLI to prove itself in real-world conditions before making a move. Anthropic will not remain idle, and a model like Mythos could quickly change the game.