GPT-5.6 Luna降價80%評測：OpenAI引爆AI價格戰 | GPT-5.6 Luna Price Cut Review: OpenAI's AI Price War
By Kit 小克 | AI Tool Observer | 2026-08-07
🇹🇼 GPT-5.6 Luna降價80%評測：OpenAI引爆AI價格戰
GPT-5.6 Luna降價80%是這週AI圈最實際的一則消息：OpenAI在7月30日無預警把Luna模型的API價格砍到只剩原本的兩成，每百萬token的綜合成本從7美元降到1.4美元。這不是單一調價,而是Anthropic、Google幾乎同時祭出對應動作後,整個「AI價格戰」正式開打的訊號。
GPT-5.6 Luna降價後多少錢？
降價後Luna輸入token每百萬只要0.2美元、輸出1.2美元,兩者合計1.4美元,比原本的7美元便宜80%。同系列的Terra模型也降了20%,但旗艦Sol維持原價不動,反而把原本的Priority Processing換成收費更高、速度快2.5倍的Fast模式。
OpenAI為什麼現在降價？
官方說法是內部代號Sol的核心重寫(kernel rewrites)把伺服成本砍了20%,把省下的錢回饋給用戶。但更現實的原因是競爭壓力:Anthropic才剛用同一個價位推出效能更強的Claude Opus 5,Google也端出主打低成本推論的Gemini 3.6 Flash和Flash-Lite,OpenAI再不降價,中小型客戶和自動化workflow用戶就有可能被搶走。加上OpenAI同期宣布月活躍用戶突破10億,維持大規模低價服務的能力,變成留住用戶的關鍵籌碼。
這波AI價格戰對開發者有什麼影響？
對正在用API做自動化、內部工具或agent workflow的開發者來說,這是實際能省錢的消息,尤其是高用量、對延遲不敏感的批次任務。
- 成本敏感應用：客服機器人、內容生成這類大量呼叫的場景,Luna現在的價格比多數同級模型便宜。
- 效能仍有取捨：Luna定位是輕量模型,複雜推理、長文件分析建議還是用Sol或對手的旗艦模型。
- 比價變成常態：Claude、Gemini、GPT-5.6三方同時在打價格戰,選型前務必自己跑基準測試,不要只看官方公布的數字。
常見問題 FAQ
Q: GPT-5.6 Luna降價後多少錢？
A: 降價後輸入token每百萬0.2美元、輸出1.2美元,合計1.4美元,比原價7美元便宜80%。
Q: 為什麼OpenAI要在這時候降價？
A: 官方說是核心重寫省下20%伺服成本,但更關鍵的是要對抗Claude Opus 5和Gemini 3.6 Flash的低價競爭。
Q: 旗艦Sol也降價了嗎？
A: 沒有,Sol價格不變,但把Priority Processing換成收費更高、速度快2.5倍的Fast模式。
Q: 這對一般開發者有實際影響嗎？
A: 有,尤其是大量呼叫API做客服、內容生成等批次任務的團隊,成本可明顯下降,但複雜推理仍建議用旗艦模型。
好不好用，試了才知道。
🇺🇸 GPT-5.6 Luna Price Cut Review: OpenAI's AI Price War
OpenAI's GPT-5.6 Luna price cut of 80% is the most practical AI story this week. On July 30, OpenAI quietly slashed Luna's API pricing to a fifth of its original cost, dropping the combined input-plus-output rate per million tokens from $7 to $1.40. It's not an isolated markdown — it's the clearest sign yet that a full-blown AI price war is underway, with Anthropic and Google moving in near lockstep.
How much did GPT-5.6 Luna's price drop?
After the cut, Luna costs $0.20 per million input tokens and $1.20 per million output tokens — $1.40 combined, down 80% from the original $7. Terra, the mid-tier model in the same family, dropped 20%. The flagship Sol model kept its price but swapped Priority Processing for a pricier Fast mode that runs 2.5x faster.
Why is OpenAI cutting prices now?
Officially, OpenAI credits kernel rewrites under the internal "Sol" effort that trimmed serving costs by 20%, passed on to customers. The more realistic driver is competitive pressure: Anthropic just launched a stronger Claude Opus 5 at the same price point, and Google shipped cost-focused Gemini 3.6 Flash and Flash-Lite models. If OpenAI didn't move, cost-sensitive startups and automation-heavy customers would drift elsewhere. OpenAI also disclosed it crossed 1 billion monthly active users around the same time — keeping pricing competitive at that scale is now central to retention.
What does this AI price war mean for developers?
If you're running automation, internal tools, or agent workflows on the API, this is real savings — especially for high-volume, latency-tolerant batch jobs.
- Cost-sensitive use cases: For high-volume tasks like customer support bots or content generation, Luna now undercuts most comparable models.
- Performance trade-offs remain: Luna is a lightweight model — complex reasoning or long-document analysis still calls for Sol or a rival flagship.
- Price comparison is now routine: With Claude, Gemini, and GPT-5.6 all cutting prices simultaneously, run your own benchmarks before switching — don't trust the headline numbers alone.
FAQ
Q: How much cheaper is GPT-5.6 Luna now?
A: $0.20/M input and $1.20/M output tokens, $1.40 combined — an 80% cut from the original $7.
Q: Why did OpenAI cut prices now?
A: Officially, cost savings from kernel rewrites; realistically, competition from Claude Opus 5 and Gemini 3.6 Flash.
Q: Did the flagship Sol model also get cheaper?
A: No — Sol's price is unchanged, but it now offers a Fast mode that's 2.5x faster at a higher price.
Q: Does this actually matter for developers?
A: Yes, especially for high-volume batch tasks like support bots or content generation. Complex reasoning still benefits from flagship models.
好不好用，試了才知道。
Sources / 資料來源
- VentureBeat: AI price wars — OpenAI cuts GPT-5.6 Luna prices by 80%
- CNBC: OpenAI cuts prices for two of its GPT-5.6 AI models
- gHacks: OpenAI reaches one billion active users and cuts GPT-5.6 Luna and Terra prices
常見問題 FAQ
GPT-5.6 Luna降價後多少錢？
降價後輸入token每百萬0.2美元、輸出1.2美元,合計1.4美元,比原價7美元便宜80%。
為什麼OpenAI要在這時候降價？
官方說是核心重寫省下20%伺服成本,但更關鍵的是要對抗Claude Opus 5和Gemini 3.6 Flash的低價競爭。
旗艦Sol也降價了嗎？
沒有,Sol價格不變,但把Priority Processing換成收費更高、速度快2.5倍的Fast模式。
這對一般開發者有實際影響嗎？
有,尤其是大量呼叫API做客服、內容生成等批次任務的團隊,成本可明顯下降,但複雜推理仍建議用旗艦模型。
延伸閱讀 / Related Articles
- CrowdStrike威脅報告評測：AI攻擊暴增89%、29分鐘破網 | CrowdStrike 2026 Threat Report: AI Attacks Surge 89%
- Big Sleep AI評測：Google揪出Chrome藏13年漏洞 | Big Sleep AI Review: Google Finds 13-Year Chrome Bug
- Claude Opus 5評測：半價戰勝Fable 5的寫程式神器 | Claude Opus 5 Review: Frontier Coding at Half the Price
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
留言
張貼留言