Gemini 3.7 Flash評測：降價五成、程式碼能力大躍進 | Gemini 3.7 Flash Review: 50% Price Cut, Coding Leap

8月 16, 2026

By Kit 小克 | AI Tool Observer | 2026-08-17

🇹🇼 Gemini 3.7 Flash評測：降價五成、程式碼能力大躍進

Gemini 3.7 Flash是Google在2026年8月13日悄悄丟出的新模型，距離上一代Gemini 3.6 Flash上市只隔三週。這次更新的重點很單純：程式碼能力大幅進步，價格卻直接砍半，主打的就是「用得起的AI寫程式助手」這個定位。

Gemini 3.7 Flash 價格砍半是真的划算嗎？

Google把Gemini 3.7 Flash的介紹性定價訂在每百萬輸入token 0.75美元、每百萬輸出token 3.75美元，效期到2026年底。對照Gemini 3.6 Flash上市時的1.50美元／7.50美元，等於直接打對折。對每天要跑上千次API呼叫的開發團隊來說，這不是行銷噱頭，是實際能砍掉一半AI帳單的數字。

程式碼能力：DeepSWE分數從49%衝到65.3%

DeepSWE v1.1：從約49%提升到65.3%，這是目前公開的軟體工程實測基準之一，模擬真實repo修bug、加功能的任務
FrontierCode 1.1 Main：從34.4%上升到43.6%
WebDev Arena Elo：從1538漲到1588，代表前端網頁產出品質也有感提升

單看數字，Gemini 3.7 Flash在16個百分點的DeepSWE進步幅度，比多數模型一次改版的升級速度快得多，這也是為什麼Google選擇只隔三週就推新版，而不是照原本節奏等Gemini Pro先出。

誰該現在就換Gemini 3.7 Flash？

如果你的產品或工作流程本來就是用Gemini Flash系列做程式碼生成、agent任務或高流量的知識工作應用，這次改版幾乎沒有理由不升級——同樣的API呼叫方式，價格更低、輸出品質更好。已經整合在Vertex AI、AI Studio和Gemini API的專案，理論上可以直接切換測試。

但如果你在乎的是最高階推理能力，Gemini 3.7 Flash仍然是「輕量級」定位，不是拿來取代Gemini Pro或Ultra。它的價值在於用更低成本處理大量、中等複雜度的程式碼與網頁開發任務，適合把它當成日常開發的主力，而不是唯一選項。

好不好用，試了才知道

🇺🇸 Gemini 3.7 Flash Review: 50% Price Cut, Coding Leap

Google quietly shipped Gemini 3.7 Flash on August 13, 2026 — just three weeks after Gemini 3.6 Flash launched. The pitch is straightforward: a big jump in coding ability paired with a 50% price cut, aimed squarely at developers who want an AI coding assistant they can actually afford to run at scale.

Is the Gemini 3.7 Flash Price Cut Actually Worth It?

Google set introductory pricing for Gemini 3.7 Flash at $0.75 per million input tokens and $3.75 per million output tokens, good through the end of 2026. Compare that to Gemini 3.6 Flash launch pricing of $1.50 / $7.50 per million tokens, and it is a literal 50% cut. For teams running thousands of API calls a day, that is not marketing spin — it is a number that can cut your AI bill in half.

Coding Benchmarks: DeepSWE Jumps from 49% to 65.3%

DeepSWE v1.1: rose from roughly 49% to 65.3%, one of the more realistic software-engineering benchmarks, simulating bug fixes and feature work in real repos
FrontierCode 1.1 Main: up from 34.4% to 43.6%
WebDev Arena Elo: climbed from 1538 to 1588, a noticeable bump in front-end code quality

A 16-point jump on DeepSWE in a single release cycle is faster than most models improve across a full generation. That is likely why Google broke its usual cadence and shipped Gemini 3.7 Flash just three weeks after 3.6, instead of waiting for the next Gemini Pro release.

Who Should Switch to Gemini 3.7 Flash Now?

If you are already using the Gemini Flash line for code generation, agent workflows, or high-volume knowledge work, there is almost no reason not to upgrade — same API surface, lower cost, better output. Projects already wired into Vertex AI, AI Studio, or the Gemini API should be able to swap the model string and test immediately.

That said, Gemini 3.7 Flash is still positioned as the lightweight tier, not a replacement for Gemini Pro or Ultra-class reasoning. Its real value is handling large volumes of medium-complexity coding and web-dev tasks cheaply — a solid daily driver, not your only option for the hardest problems.

好不好用，試了才知道

Sources / 資料來源

延伸閱讀 / Related Articles

AI 工具觀察站 — 每日精選 AI Agent 與工具趨勢
AI Tool Observer — Daily curated AI Agent & tool trends

搜尋此網誌

AI小貼士