GEMINI 3 PREVIEW1M CONTEXTMULTIMODALCONTEXT CACHINGBATCH + FLEX -50%
Gemini 3.1 Pro Preview API Pricing
Google's current Gemini 3.1 Pro preview tier for multimodal reasoning and agentic work: $2/M input, $12/M output, and $0.20/M cached input for prompts up to 200K tokens. Pulled directly from ai.google.dev.
Input - per 1M tokens
$2.00/M
Tier 1 <=200K prompt standard
Output - per 1M tokens
$12.00/M
Tier 2 output is $18/M standard
Cached input - 90% off
$0.20/M
Cache plus storage fee -90%
Effective - agentic blend
$1.44/M
92/8 split - 82% cache
§ 01 / TERMINAL
Gemini 3.1 Pro Preview API cost calculator.
Live calculator pre-loaded with current Gemini 3.1 Pro Preview tier-1 rates. For prompts above 200K tokens, Google lists higher tier-2 input, cache, and output prices.
$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS
Gemini 3.1 Pro Preview cost on real workloads.
CODING
Vibe-coding feature
$0.138/task
LONG CONTEXT
Document pack analysis
$0.165/pack
MULTIMODAL
Video + docs briefing
$0.052/brief
AGENT
Research agent loop
$0.183/loop
RAG
Large RAG synthesis
$0.223/answer
Estimate · gemini-tokenizer-estimate · ≈3.85 chars/token Auto-counts as you type
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
Characters 483
Words 74
Tokens (estimated) 125 tokens
Cost as input · uncached $0.00025 USD
Cost as output · uncached $0.0015 USD
Cost as cached input $0.00003 USD
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Gemini 3.1 Pro Preview Current | $2.00 cache $0.20 | $12.00 | $1.44 agentic 92/8 | 1M | Google frontier preview |
| Gemini 3 Flash Preview | $0.50 cache $0.05 | $3.00 | $0.361 −4.0× | 1M | Cheaper Gemini 3 preview |
| Gemini 3.5 Flash-Lite | $0.30 cache $0.03 | $2.50 | $0.272 −5.3× | 1M | Google's high-volume Lite tier |
| Gemini 2.5 Pro | $1.25 cache $0.125 | $10.00 | $1.10 deprecated · −24% | 1M | Stable Gemini 2.5 Pro |
| Gemini 2.5 Flash | $0.30 cache $0.03 | $2.50 | $0.272 deprecated · −5.3× | 1M | Best price-performance Gemini 2.5 |
| GPT-6.1 Sol | $2.00 cache $0.10 | $10.00 | $1.29 −11% | 1.05M | OpenAI's balanced model for coding and agents |
| GPT-6.1 Terra | $1.00 cache $0.05 | $5.00 | $0.645 −2.2× | 1.05M | OpenAI's fast, efficient GPT-6 tier |
Gemini 3.1 Pro Preview has published scores on 14 benchmarks we track; its best placing is #7 of 99 on ARC-AGI-1. 3 of these scores come from an independent leaderboard; the rest are the lab's own reports.
| Benchmark | Score | Rank | Leader | Source |
|---|---|---|---|---|
| ARC-AGI-1 Reasoning | 98% ARC Prize verified, best result | #7 of 99 | 6 tied · 98.5% | Official leaderboard ARC Prize Foundation |
| ARC-AGI-2 Reasoning | 77.1% ARC Prize verified, best result | #20 of 105 | GPT-6 Astra · 95% | Official leaderboard ARC Prize Foundation |
| GPQA Diamond Expert knowledge | 94.3% no tools, thinking high | #3 of 11 | GPT-6 Astra · 96% | Vendor-reported Google DeepMind |
| MRCR v2 Long context · 8-needle, up to 128K | 84.9% 128K average | #3 of 7 | Gemini 3.7 Flash · 97% | Vendor-reported Google DeepMind |
| SWE-Bench Pro Coding · Public set (731 tasks) | 54.2% public set | #3 of 6 | Gemini 3.6 Flash · 58.7% | Vendor-reported Google DeepMind |
| MLE-bench AI research | 42.6% as reported | #3 of 5 | Gemini 3.6 Flash · 63.9% | Vendor-reported Google DeepMind |
| ARC-AGI-3 Reasoning · Standard harness | 0.4% ARC Prize verified, best result | #15 of 24 | GPT-6 Astra · 62.7% | Official leaderboard ARC Prize Foundation |
| OSWorld-Verified Agents and computer use | 76.2% as reported | #6 of 8 | Kimi K3 · 84.8% | Vendor-reported Google DeepMind |
| CharXiv Reasoning Vision and documents · No tools | 83.3% no tools | #7 of 9 | Kimi K3.1 Pro · 89% | Vendor-reported Google DeepMind |
| APEX-Agents Professional work | 33.5% thinking high | #5 of 6 | Kimi K3.1 Pro · 69.4% | Vendor-reported Google DeepMind |
| BrowseComp Agents and computer use | 85.9% search, Python and browse, thinking high | #5 of 6 | Kimi K3.1 Pro · 92.1% | Vendor-reported Google DeepMind |
| Terminal-Bench 2 Agents and computer use · Terminal-Bench 2.1 | 73.8% Terminus-2 harness | #15 of 17 | DeepSeek V4.1 Flash · 90.6% | Vendor-reported Google DeepMind |
| Humanity's Last Exam Expert knowledge · With tools | 51.4% search and code, thinking high | #18 of 20 | Claude Fable 5.2 · 74.2% | Vendor-reported Google DeepMind |
| DeepSWE Coding | 12% as reported | #27 of 27 | Gemini 4 Argon · 77.9% | Vendor-reported Google DeepMind |
Gemini 3.1 Pro Preview API pricing FAQ.
Practical Gemini 3.1 Pro Preview pricing questions, with Google's tiered prompt prices separated from workload assumptions.
Q · 01 What is Gemini 3.1 Pro Preview's standard API price? +
Google lists
gemini-3.1-pro-preview at $2/M input, $0.20/M cached input, and $12/M output for prompts up to 200K tokens. For prompts above 200K, the table lists $4/M input, $0.40/M cached input, and $18/M output.Q · 02 Does output pricing include thinking tokens? +
Yes. Google's pricing page labels output as
Output price (including thinking tokens). This page therefore treats all generated reasoning and answer tokens as part of the published output rate.Q · 03 How much do Batch and Flex cost? +
Google lists Batch and Flex at
$1/M input and $6/M output for prompts up to 200K, with $2/M input and $9/M output above 200K. Context caching in those sections is shown at the same cache rate as Standard.Q · 04 Is this a preview model? +
Yes. Google's models page labels Gemini 3.1 Pro as
Preview. Preview models may have more restrictive limits and can change before stable release.Q · 05 What about Google Search grounding costs? +
For Gemini 3 models, Google lists 5,000 grounded prompts per month free, shared across Gemini 3, then
$14 / 1,000 search queries. Tool charges are separate from token prices.Q · 06 How accurate is the tokenizer estimate? +
The widget uses
4.0 characters per token as a Gemini planning estimate. Exact billing can differ by language, media inputs, tool calls, and how Google tokenizes multimodal content.