Gemini 2.5 Flash → Llama 3.3 70B Instruct: cost per request on six tasks
Llama 3.3 70B Instruct costs between 47% less and 117% more than Gemini 2.5 Flash per request on six common tasks. It costs no more per result that works if its results work first time at least 54% as often as Gemini 2.5 Flash's on translation, and 2.2 times as often on search re-ranking. Google has announced Gemini 2.5 Flash's API shutdown for 2026-12-01.
List prices per million tokens
| Model | Input | Cached input | Output |
|---|---|---|---|
| Gemini 2.5 Flash | $0.30 | $0.03 | $2.50 |
| Llama 3.3 70B Instruct | $0.71 | $0.71 | $0.71 |
Prices checked 2026-11-12.
Cost per 1,000 requests
| Task | Gemini 2.5 Flash | Llama 3.3 70B Instruct | Change | Cheaper per result that works if it works first time at least |
|---|---|---|---|---|
| Support bot answerInstructions, three help-centre passages and the customer's question in; a short answer out. | $1.65 | $2.34 | 42% more | 42% more often than Gemini 2.5 Flash |
| Invoice field extractionInstructions and one invoice's text in; about 15 fields as JSON out. | $1.38 | $1.95 | 42% more | 42% more often than Gemini 2.5 Flash |
| Lead scoringScoring rules and one lead's form answers in; a score and a one-line reason out. | $0.61 | $0.92 | 51% more | 52% more often than Gemini 2.5 Flash |
| Email reply draftInstructions and the email thread so far in; a reply of about 300 words out. | $1.45 | $1.35 | 7% less | 94% as often as Gemini 2.5 Flash |
| TranslationInstructions and about 500 words of text in; the same text in another language out. | $2.27 | $1.21 | 47% less | 54% as often as Gemini 2.5 Flash |
| Search re-rankingA search query and 20 result snippets in; the ids of the best five, in order, out. | $1.33 | $2.88 | 2.2 times as much | 2.2 times as often as Gemini 2.5 Flash |
Token counts are typical sizes, not measurements.
Gemini 2.5 Flash is being shut down
Google has announced Gemini 2.5 Flash's API shutdown for 2026-12-01, with gemini-3.5-flash-lite (requests auto-route, billed at 3.5 Flash-Lite rates) as the replacement (source).
Test it on your own prompt
The model switch test runs your prompt on both models side by side with your own API key, in your browser.