Google ships Gemini 3.7 Flash with coding gains and 50% price cut
Google has shipped Gemini 3.7 Flash just three weeks after its predecessor, targeting coding and agent workloads with significant benchmark gains and a 50% price cut. On FrontierCode the model scores 43.6%, up from 34.4%, and on DeepSWE it reaches 65.3% versus 49.0% for 3.6 Flash. Launch pricing is $0.75 per million input tokens and $3.75 per million output tokens, half the cost of 3.6 Flash. The model is already integrated into Devin and the Gemini Spark tier, and Google claims it beats Claude Sonnet 5 and GPT-5.6 Terra on its own benchmarks. This rapid iteration and aggressive undercutting signal a shift from capability competition to cost-performance dominance. However, some sources raise questions about the model's availability and official status, so builders should verify details.
Google ships Gemini 3.7 Flash with coding gains and 50% price cut
Google shipped Gemini 3.7 Flash just three weeks after 3.6 Flash. On the FrontierCode benchmark, the model scores 43.6 percent, up from 34.4 percent. On DeepSWE, it hits 65.3 percent versus 49.0 percent. Launch pricing sits at $0.75 per million input tokens and $3.75 per million output tokens, 50 percent cheaper than 3.6 Flash at launch.
Key takeaway
Google is weaponizing price and iteration speed to dominate the coding model tier, making cost-performance the primary differentiator and forcing rivals to justify premium pricing.
What happened
Google shipped Gemini 3.7 Flash just three weeks after 3.6 Flash, positioning it as its most capable workhorse for coding and agents. The company reports sizeable gains on coding benchmarks: FrontierCode rose to 43.6% from 34.4%, and DeepSWE to 65.3% from 49.0%.
Launch pricing is $0.75 per million input tokens and $3.75 per million output tokens, 50% cheaper than 3.6 Flash at launch. The model is already available in Devin Desktop and CLI, and integrated into the Gemini Spark tier for Google AI Pro and Ultra, with introductory pricing through year-end.
Evidence
Gemini 3.7 Flash scores 43.6% on FrontierCode, up from 34.4% for 3.6 Flash.
The Decoder · attributed
On the FrontierCode benchmark, the model scores 43.6 percent, up from 34.4 percent.
Launch pricing is $0.75 per million input tokens and $3.75 per million output tokens, 50% cheaper than 3.6 Flash.
The Decoder · attributed
Launch pricing sits at $0.75 per million input tokens and $3.75 per million output tokens, 50 percent cheaper than 3.6 Flash at launch.
Devin has integrated Gemini 3.7 Flash into its Desktop and CLI, claiming Claude Sonnet 5-level performance at less than half the cost.
Windsurf Blog · attributed
Gemini 3.7 Flash is now live in Devin Desktop and Devin CLI, reaching Claude Sonnet 5-level performance at less than half the cost.
Google claims superior performance against Claude Sonnet 5 and GPT-5.6 Terra on its own benchmarks.
The Decoder · attributed
according to Google's own measurements, that puts it ahead of both Claude Sonnet 5 and GPT-5.6 Terra.
Gemini 3.7 Flash is a refinement with algorithmic improvements and a 1M-token context window.
MarkTechPost · attributed
Google has released Gemini 3.7 Flash, a refinement of Gemini 3.6 Flash with algorithmic improvements to its reasoning core. It handles text, images, audio, and video across a 1M-token context window with 64K-token outpu…
The model is integrated into Gemini Spark tier for Google AI Pro and Ultra.
Hacker News AI · attributed
The update is significant because it is being integrated into the Gemini Spark tier for Google AI Pro and Ult…
Why it matters
Builders and operators running high-volume agent or coding workloads can materially cut inference costs by switching to Gemini 3.7 Flash, potentially shifting default choices from premium models to cost-optimized tiers.
Limits and uncertainties
One Hacker News AI source speculates the model may be erroneous due to data extraction glitches, while another confirms integration, indicating conflicting information.
Benchmark claims are based on Google's own measurements and have not been independently verified.
The 50% price cut is introductory and may change after the end of the year.
Practical implications
Operators should re-run cost analysis for coding and agent workloads using the new pricing to capture immediate savings.
Devin's integration suggests agents can leverage the model for high-volume tasks without sacrificing quality, reducing operational expenses.
Given rapid iteration, builders should architect for model flexibility to switch to newer Flash versions as they emerge.
What to watch
Independent benchmark results comparing Gemini 3.7 Flash to Claude Sonnet 5 and GPT-5.6 Terra.
Whether competitors respond with price cuts or capability improvements to counter Google's move.
Adoption rates and real-world performance reports from Devin and other agents using the model.
