Gemini API pricing is a per-million-token rate for input, output and cached input, but your bill depends on how many tokens each task uses. Gemini 3.8 Flash and Gemini 3.1 Pro Preview both reason by default (my data lists "always"), and reasoning tokens are billed as output. In my measurements, Gemini 3.8 Flash was cheaper per task than 3.1 Pro Preview on every task I ran, but it was not the cheapest option against OpenAI and Anthropic counterparts.
List prices vs. cost per task
Per-token list prices are a starting point, not the answer. A model with a low input price can still be expensive if it writes long reasoning traces. My methodology counts reasoning tokens as output tokens, so they show up in the bill.
List price per 1M tokens for the Google models and their closest OpenAI and Anthropic counterparts:
| Model | Input | Output | Cached input | LMArena Text |
|---|---|---|---|---|
| Gemini 3.8 Flash | $0.75 | $3.75 | $0.075 | 1496 |
| GPT-5.4 Mini | $0.75 | $4.50 | $0.075 | 1447 |
| Gemini 3.1 Pro Preview | $2.00 | $12.00 | $0.20 | 1487 |
| GPT-6.1 Sol | $2.00 | $10.00 | $0.10 | 1476 |
| Claude Sonnet 5.5 | $2.00 | $10.00 | $0.10 | 1476 |
| Claude Opus 5.5 | $4.00 | $20.00 | $0.20 | 1507 |
Gemini 3.8 Flash matches GPT-5.4 Mini on input and cached input, and its output is cheaper. Gemini 3.1 Pro Preview has a higher output price than GPT-6.1 Sol or Claude Sonnet 5.5. Claude Opus 5.5 is the most expensive of the group per token.
Reasoning is on by default
Both models have reasoning always on in my data. Thinking tokens are billed at the output rate, so token counts matter as much as list price.
Reasoning loads varied by task. On the spreadsheet task, Gemini 3.8 Flash used 7,100 reasoning tokens and Gemini 3.1 Pro Preview used 4,358. On code review, Gemini 3.8 Flash used 2,074 output tokens and no reasoning tokens in my run, while 3.1 Pro Preview used 1,368 output and 7,256 reasoning tokens. That difference explains most of the gap in the code review bill.
Cost per task for Gemini 3.8 Flash
These are measured results: one run per task, default settings, via OpenRouter, output capped at 16,000 tokens.
| Task | Gemini 3.8 Flash | Median of 326 models |
|---|---|---|
| Answer a chat question | 0.61¢ ($6.14 per 1,000) | 0.24¢ |
| Summarize a 20-page PDF | 1.4¢ ($14.24 per 1,000) | 0.83¢ |
| Analyze a 200-row spreadsheet | 3.5¢ ($35.17 per 1,000) | 2.1¢ |
| Review 500 lines of code | 1.2¢ ($11.68 per 1,000) | 1.4¢ |
Gemini 3.8 Flash is above the median on three tasks and below it on code review. On chat it is close to GPT-6.1 Sol at 0.67¢.
Pro tier: Gemini vs. OpenAI vs. Anthropic
Cost per task for Gemini 3.1 Pro Preview and its closest rivals:
| Task | Gemini 3.1 Pro Preview | GPT-6.1 Sol | Claude Sonnet 5.5 | Claude Opus 5.5 |
|---|---|---|---|---|
| Chat | 2.1¢ | 0.67¢ | 1.3¢ | 2.5¢ |
| PDF summary | 4.6¢ | 3.1¢ | 4.5¢ | 8.9¢ |
| Spreadsheet | 7.5¢ | 8.4¢ | 17¢ | 34¢ |
| Code review | 11¢ | 4.9¢ | 8.0¢ | 16¢ |
GPT-6.1 Sol was cheaper than Gemini 3.1 Pro Preview on chat, PDF and code review. On the spreadsheet task, Gemini 3.1 Pro Preview was cheaper than GPT-6.1 Sol and both Claude models. Claude Sonnet 5.5 and Claude Opus 5.5 hit the 16,000-token cap on that task, which inflated their bills.
For the live rankings, see the chat task page, PDF summary, spreadsheet analysis and code review.
| # | Model | Cost per task | Text score | Tokens in / out | Basis | |
|---|---|---|---|---|---|---|
| 1 |  GLM 5.3 FlashZ.ai | 0.04¢$0.36 / 1,000 | 1474 | 48 / 1,782 | measured best value | |
| 2 |  GPT-6 LunaOpenAI | 0.06¢$0.59 / 1,000 | 1443 | 45 / 1,179 | measured | |
| 3 |  Claude Haiku 5.5Anthropic | 0.07¢$0.66 / 1,000 | - | 61 / 1,299 | measured | |
| 4 |  Qwen3.8 FlashAlibaba Qwen | 0.08¢$0.76 / 1,000 | - | 88 / 1,598 | measured | |
| 5 |  Mistral Small 4Mistral | 0.08¢$0.83 / 1,000 | - | 40 / 1,377 | estimated | |
| 6 |  MiMo-V2.6-ProXiaomi | 0.09¢$0.90 / 1,000 | 1479 | 206 / 933 | estimated best value | |
| 7 |  DeepSeek V4.1 FlashDeepSeek | 0.09¢$0.91 / 1,000 | 1475 | 65 / 6,320 | measured | |
| 8 |  Llama 4 MaverickMeta | 0.11¢$1.06 / 1,000 | 1327 | 87 / 1,592 | estimated | |
| 9 |  MiniMax M3MiniMax | 0.12¢$1.18 / 1,000 | 1441 | 205 / 929 | measured | |
| 10 |  Command A+Cohere | 0.15¢$1.46 / 1,000 | - | 205 / 930 | estimated |
Cheaper Google options
Gemini 3.1 Flash Lite cost 0.25¢ on the chat task and Gemini 3.5 Flash Lite cost 0.41¢, both estimated. On code review, Gemini 3.1 Flash Lite was estimated at 0.44¢, well under Gemini 3.8 Flash at 1.2¢. The cheapest chat answer in my set was GPT-6 Luna at 0.06¢ (measured).
The tokenizer effect
The same English document produces different token counts by provider. Against OpenAI at 1.00, Anthropic comes in at 1.39 and Google at 0.98. A higher token count raises the cost of the same input.
Per-token price alone can mislead. Claude Haiku 5.5 lists at $0.10 input and $0.50 output, well below Gemini 3.8 Flash, and it was cheaper on the PDF task (0.22¢ vs 1.4¢). On code review it cost 0.66¢ against 1.2¢. Token counts and reasoning length decide the rest.
Gemini 3.8 Flash pricing is promotional
Google lists Gemini 3.8 Flash and 3.6 Flash at $0.75 input and $3.75 output per 1M tokens until December 31, 2026. The official page says the price doubles on January 1, 2027. Every Gemini 3.8 Flash figure here reflects the promotional rate. For 2027 budgets, use the post-promotion price from Google's pricing page.
Estimate your own cost
Multiply list prices by your task's token counts, including reasoning tokens. The calculator does this for you. Check context size and reasoning mode on the Gemini 3.8 Flash model page before choosing.
FAQ
Is Gemini 3.8 Flash cheaper than GPT-6.1 Sol per task?
In my data, yes on all four tasks: 0.61¢ vs 0.67¢ on chat, 1.4¢ vs 3.1¢ on PDF, 3.5¢ vs 8.4¢ on the spreadsheet and 1.2¢ vs 4.9¢ on code review. This is at the promotional Gemini price.
Why does Gemini 3.1 Pro Preview cost more per task than Gemini 3.8 Flash?
Its list price is higher at $2.00 input and $12.00 output, and on code review it also used 7,256 reasoning tokens against none for Gemini 3.8 Flash. That put code review at 11¢ vs 1.2¢.
Will Gemini 3.8 Flash stay at the current price?
No. Google lists $0.75 input and $3.75 output until December 31, 2026, and says the price doubles on January 1, 2027.