Google launches Gemini 3.7 Flash with 50% price cut and 65% DeepSWE score
Google has introduced Gemini 3.7 Flash, a rapid iteration on its Flash line, arriving just three weeks after 3.6 Flash. The new model is available on APIs, AI Studio, and Antigravity, with broader rollout planned. Notably, it is 50% cheaper than its predecessor, positioning it as a cost-efficient workhorse for developers. On the DeepSWE V1.1 benchmark, it scored 65%, approaching the performance of GPT-5.6 Terra, signaling competitive gains in coding and agentic tasks. This aggressive pricing and swift release cadence could pressure rivals in the API market. However, the DeepSWE score is a single benchmark; real-world agentic reliability remains unverified, and historical performance may not directly translate to production outcomes. Developers should test it against their specific workloads.
Google launches Gemini 3.7 Flash with 50% price cut and 65% DeepSWE score
Google announced Gemini 3.7 Flash, now available on APIs, AI Studio, and Antigravity, with rollout to other products. It ships 3 weeks after 3.6 Flash, is 50% cheaper than 3.6 Flash, and scored 65% on DeepSWE V1.1 close to GPT-5.6 Terra performance.
Key takeaway
Gemini 3.7 Flash delivers a 50% price cut and a 65% DeepSWE score close to GPT-5.6 Terra, signaling a cost-performance leap for agentic coding workloads.
What happened
According to the announcement on X, Google unveiled Gemini 3.7 Flash, making it available on APIs, AI Studio, and Antigravity, with plans to roll out to other products. The release comes just three weeks after the 3.6 Flash launch, underscoring a fast iteration cycle.
The new model is priced 50% lower than Gemini 3.6 Flash and achieved a 65% score on the DeepSWE V1.1 benchmark, which the announcement notes is close to GPT-5.6 Terra performance. This indicates notable progress in coding and agentic capabilities at a reduced cost.
Evidence
Gemini 3.7 Flash is 50% cheaper than Gemini 3.6 Flash.
X · attributed
Gemini 3.7 Flash is 50% cheaper than Gemini 3.6 Flash.
Gemini 3.7 Flash scored 65% on DeepSWE V1.1, close to GPT-5.6 Terra performance.
X · attributed
It scored 65% on DeepSWE V1.1, getting close to GPT-5.6 Terra performance.
The release follows 3.6 Flash by three weeks.
X · attributed
And now just 3 weeks after launching 3.6 Flash
Why it matters
The 50% price reduction combined with near-frontier coding performance on DeepSWE could force competitors to accelerate iteration cycles and rethink API pricing, directly impacting developer tooling choices and AI infrastructure costs.
Limits and uncertainties
The DeepSWE score is a single benchmark; no further evaluation details are provided in the announcement.
The rollout is initially limited to APIs, AI Studio, and Antigravity; other products are unspecified.
Practical implications
Developers should benchmark Gemini 3.7 Flash against their own agentic workflows to validate the DeepSWE gains in production.
The 50% price reduction could lower API costs for high-volume coding tasks, so teams should update cost models accordingly.
What to watch
Independent third-party results on agentic coding suites, such as SWE-bench or similar, to confirm DeepSWE parity.
Google's release cadence for subsequent Flash updates to see if the three-week cycle continues.
Original reporting: Our Flash models are workhorses that offer performance at a great price. So, we're shipping updates fast to get them in developers’ hands. And now just 3 weeks after launching 3.6
