Gemini 3.7 Flash — three weeks after 3.6, half the intro price
Google’s coding and agents workhorse jumps again: stronger SWE / web / document benches vs 3.6 Flash, Spark switches over, and intro API pricing at $0.75 / $3.75 through year-end.
Read time: ~8 minutes
Gemini 3.7 Flash is live — Google’s “most intelligent workhorse yet for coding and agents,” landing about three weeks after Gemini 3.6 Flash. That cadence matters: Flash used to feel like a seasonal refresh; this is a sprint. Google says the jump comes from developer feedback plus algorithmic work they expect to fold into later models.
The other headline is price. Introductory API rates are half of the original 3.6 Flash list: $0.75 / $3.75 per million input/output tokens through 31 December 2026. On 1 January 2027 they snap back to $1.50 / $7.50 — the same list we covered for 3.6.
What Google claims improved
Versus 3.6 Flash, Google highlights debugging / issue resolution, higher first-pass code accuracy, better production-ready generation, stronger web UI adherence (screenshot → layout), and denser document / workflow reasoning (finance, law, biosciences). DX side: clearer intent clarification, more diligent multi-step planning and tool calls, fewer retries.
Gemini Spark — the 24/7 personal agent for Google AI Pro and Ultra in supported countries — starts using 3.7 Flash for Workspace-heavy knowledge work (consolidate files, draft email, update status docs).
Pricing: intro window vs list
| Model | Input / 1M | Output / 1M | Notes |
|---|---|---|---|
| Gemini 3.7 Flash (intro) | $0.75 | $3.75 | Through 31 Dec 2026 |
| Gemini 3.7 Flash (from 2027) | $1.50 | $7.50 | Same list as today’s 3.6 Flash |
| Gemini 3.6 Flash | $1.50 | $7.50 | Our July write-up |
| Grok 4.5 / 4.6 list | $2 | $6 | Cursor / SpaceXAI; Fast tiers higher |
| GPT-5.6 Luna | $1 | $6 | Budget OpenAI tier |
If you already pay for Gemini API usage, pair this with Google’s Postpay → Prepay move (deadline 14 Sep 2026 on AI Studio billing). Free-tier-only builders can ignore Prepay; paid agent loops cannot.
Comparison sheet: 3.7 vs 3.6 (Google’s numbers)
These are Google-published deltas from the Aug 13 introduction. Treat them as vendor benches — useful directionally, not as a blended “best model” score (same honesty bar we use on VCL VibeBench).
| Benchmark | 3.7 Flash | 3.6 Flash |
|---|---|---|
| FrontierCode 1.1 Main | 43.6% | 34.4% |
| DeepSWE v1.1 | 65.3% | 49.0%was ~49% on our 3.6 post (DeepSWE) |
| WebDev Arena (Elo) | 1588 | 1538 |
| GDP.pdf | 34.0% | 22.0% |
| AutomationBench | 30.4% | 17.0% |
Read: clear step up inside the Flash line on coding, web layout, long documents, and enterprise workflow automation. Cross-vendor peaks (Fable / Opus / Grok on other harnesses) still need a same-prompt check — don’t crown a king from one vendor chart.
How this fits prior VCL posts
- Gemini 3.6 Flash — the prior workhorse drop (~3 weeks earlier), Flash-Lite, and why Pro was still waiting.
- Free AI API keys (2026) — AI Studio starter key; EU free-tier availability caveats still apply.
- 5 free Google AI tools — Antigravity / Flow surfaces that sit next to the Gemini API.
- Muse Code vs Claude / Cursor pricing — mid-range price band check (3.6 Flash sat at $1.50/$7.50; 3.7 intro undercuts that hard).
- Grok 4.5 in Cursor — still the co-trained SpaceXAI lane; 4.6 is the current Cursor bump (covered in this issue’s newsletter).
Who should switch
- Default to 3.7 Flash for new Google agent / coding loops while the intro price holds.
- Keep 3.6 only if a pin/regression suite still depends on it — otherwise migrate the model id.
- Paid API users: set Prepay before 14 Sep; free-tier-only can wait.
- Stay multi-model in Cursor — Gemini when Google wins the task, Grok / Claude / GPT when they don’t.
Bottom line
Three weeks after 3.6, Google shipped a sharper Flash and temporarily halved the list price. That is the story for vibe coders this week: try 3.7 on a real agent loop, watch the intro window, and don’t sleep on the Prepay deadline if you already burn paid Gemini tokens.
Sources: Google Gemini blog — Introducing Gemini 3.7 Flash (Aug 13, 2026); prior VCL coverage of 3.6 Flash. Vendor harnesses differ — don’t treat the sheet as a single league table.
Disclosure: Some outbound links elsewhere on the site are referral or partner links (see Tools). Cursor homepage links are non-referral after the referral program ended.
Questions? Get in touch — or subscribe for the next AI news post.