Google Gemini API Pricing
Complete pricing breakdown for selected Gemini models — including the Gemini 4 Argon flagship, Gemini 3.8 Flash, and the budget-friendly 2.5 Flash-Lite
Gemini 4 Argon: generally available
Checked October 14, 2026. GA on the Gemini Developer API since October 13, 2026.
Google announced Gemini 4 Argon on September 30 with restricted Fairwind access for selected cyber defenders, then marked gemini-4-argon generally available in its changelog on October 13, 2026. It is Google's flagship reasoning model: a unified 1M-token input/output limit, text/image/video/audio/PDF input, text output, low/medium/high thinking, and tools including function calling, code execution, search and Maps grounding, URL context and file search (computer use is Preview; the Live API, audio generation and image generation are not supported). Implicit caching only — explicit caching is not supported.
Argon is not available on the Free tier or the Google AI Pro plan; paid-tier Gemini API and Google AI Studio access apply. Its 1,048,576-token budget is shared by input and output. Foreground Interactions requests, streaming or not, stop at 65,536; background execution with an explicit higher limit manages continuation hops. GenerateContent client-side continuation supports 64K per stream or 32K per non-streaming request. Use the current Google GenAI SDK or a controlled continuation loop, and check framework output caps. Official long-decode limits.
Migration requires more than a model-ID swap: explicit cached-content creation is unsupported; implicit caching starts at 4,096 input tokens. Use thinking_level, remove ignored sampling parameters and test tools, multi-turn state and timeouts. The calculator switches to the scheduled December rates after November 30 without rewriting the stored introductory prices. Google's migration checklist. Details checked October 15, 2026.
| Mode | Introductory (through Nov 30, 2026) | Standard (from Dec 1, 2026) |
|---|---|---|
| Standard input, per 1M tokens | $2.00 | $4.00 |
| Standard output, per 1M tokens (includes thinking) | $10.00 | $20.00 |
| Context caching, per 1M tokens | $0.10 | $0.20 |
| Batch input / output | $1.00 / $5.00 | $2.00 / $10.00 |
| Flex input / output | $1.00 / $5.00 | $2.00 / $10.00 |
| Priority input / output | $3.60 / $18.00 | $7.20 / $36.00 |
These are current public API prices — Argon is now included in the Gemini pricing table and the cost calculator on this page. Standard rates double on December 1, 2026; budget with the introductory rate only through November 30.
Sources: Google's Gemini API changelog (Oct 13 GA entry), Google's API pricing page (updated Oct 13, 2026) and the Argon model page. The machine-readable record is in the main models section of our pricing dataset.
Image and managed-agent migration status
Checked October 12, 2026: Google’s pricing page says gemini-2.5-flash-image shut down October 2, while its lifecycle table gives March 15, 2027. We therefore mark current callability uncertain, not definitively active or discontinued; no authenticated endpoint test was performed. Plan an explicit migration rather than relying on this disputed fallback. Image pricing remains separate from the text calculator.
antigravity-preview-05-2026 has its documented October 5 shutdown; its replacement is antigravity-preview-09-2026. Remote callers reading final output can update the ID, while local tool dispatch and function_call parsers must migrate parameter names and line-range editing. Follow Google's migration notes.
deep-research-pro-preview-12-2025 is deprecated and scheduled to shut down October 23, 2026. Use the supported deep-research-preview-04-2026 or deep-research-max-preview-04-2026 agent in the Interactions API's agent parameter, not a fabricated text-model identifier. Official migration notice.
Gemini API Pricing
Lifecycle checked October 12, 2026: gemini-3.7-flash and gemini-3.5-flash are deprecated aliases routing to 3.8 and 3.6 respectively; Google has not announced their shutdown dates. Their table rows retain historical prices, while the calculator resolves the current replacement. Official changelog.
Nano Banana 2.1 (gemini-nano-banana-2.1, GA) is an image model, not a text-model calculator substitute. Standard text/image/video input is $1.50/M; text/thinking output is $7.50/M and image output $30/M. Image-output cost depends on resolution: 1K $0.0336, 2K $0.0504, 4K approximately $0.113 per image. Batch rates are half these token rates. This is not a universal 50% reduction across all costs versus its predecessor. See Google pricing and the media records in our pricing API.
Selected Gemini text-token rates. See separate audio/video billing records; this table is not a universal modality price list.
| Model | Input (per 1M) | Output (per 1M) | Context | Tier |
|---|---|---|---|---|
| Gemini 4 Argon (GA) flagship | $2.00 thru Nov 30 → $4.00 | $10.00 thru Nov 30 → $20.00 | 1M | Premium |
| Gemini 3.8 Flash (GA) | $0.75 | $3.75 | 1M | Introductory through Dec 31, 2026 |
| Gemini 3.7 Flash Deprecated alias → 3.8 Flash | $0.75 promo | $3.75 promo | 1M | Budget |
| Gemini 3.6 Flash | $0.75 promo | $3.75 promo | 1M | Budget |
| Gemini 3.5 Flash | $1.50 | $9.00 | 1M | Mid |
| Gemini 3.5 Flash-Lite NEW | $0.30 | $2.50 | 1M | Budget |
| Gemini 3.1 Pro | $2.00 | $12.00 | 1M | Mid |
| Gemini 3.1 Flash-Lite | $0.25 | $1.50 | 1M | Budget |
| Gemini 3 Flash | $0.50 | $3.00 | 1M | Budget |
| Gemini 2.5 Pro | $1.25 | $10.00 | 1M | Mid |
| Gemini 2.5 Flash | $0.30 | $2.50 | 1M | Mid |
| Gemini 2.5 Flash-Lite | $0.10 | $0.40 | 1M | Budget |
Gemini 3.6-3.8 introductory billing checked Oct 14, 2026; older model rows retain their previous dataset verification dates. Free-tier availability and limits are model- and mode-specific; do not assume all models or processing modes are free.
Gemini 3.8 Flash billing dates and modes
Direct Gemini API rates checked September 30, 2026. Through December 31: $0.75/M input, $0.075/M cached input and $3.75/M output including thinking tokens. From January 1, 2027: $1.50/$0.15/$7.50. Cached storage is separate: $0.50/M tokens/hour now, $1/M/hour from January 1. Batch and Flex currently cost $0.375/$1.875 input/output; Priority costs $1.35/$6.75. Those input/output mode rates also double January 1. Free tiers and data-use terms differ from paid tiers; Vertex or third-party hosts may bill differently.
Google's first-party pricing. Use the cost calculator for base token estimates; mode, storage and grounding charges require separate budgeting.
Why Gemini Stands Out
Three reasons Google's API is hard to beat.
1M Context Window
Several Gemini models support up to 1M input context. This is not unique: current OpenAI GPT-6 supports 1.05M. Context capacity, output limits and long-context billing must be compared separately.
Cheapest Premium Model
Gemini rates vary by model, context and processing mode. Compare total task cost and evaluated quality rather than treating legacy token prices as equivalent intelligence.
Free Tier Available
Google offers a generous free tier for both Gemini 2.5 Pro and Gemini 2.5 Flash-Lite, letting you experiment and prototype without spending a dime. Perfect for side projects and evaluation.
Which Gemini Model Should You Use?
gemini-3.8-flash for a deliberate current choice. The deprecated 3.7 alias now routes here; compare current launch/list dates and billing modes above rather than buying on historical 3.7 rates.Gemini Cost Calculator
Estimate your monthly spend with Google Gemini models.
Monthly Cost Estimate
Based on published API pricing. Actual costs may vary.
Compare Gemini with Other Providers
Use the model comparison tool for current provider-specific rates. Match processing mode, context size and promotional eligibility before comparing; an older-model price snapshot is not a current provider-wide benchmark.
See how Gemini models stack up against OpenAI, Anthropic, and more
Deep-Dive Comparisons
Read our head-to-head analyses of Gemini vs the competition.
Historical Gemini 3.5 Flash vs Claude Haiku 4.5 comparison
Historical rates: Gemini 3.5 Flash was $1.50/$9 and Haiku 4.5 $1/$5 per million tokens. The Gemini alias now routes to 3.6 Flash; use current model pricing instead of these historical percentages.
Claude Opus 4.8 vs Gemini 3.1 Pro
Two 1M-context flagship models — Gemini is 52% cheaper on output. Full pricing and use-case comparison.
Gemini 2.5 Pro vs GPT-4o
A comparison of these earlier-generation models, not a comparison of today's newest provider offerings.
OpenAI vs Google: Full Provider Comparison
We compare every model, every tier, and every use case across OpenAI and Google's AI lineups.
Claude vs Gemini: Which Is Better Value?
Anthropic's Claude models against Google's Gemini — pricing, context limits, and real-world performance.
Calculate Your Google AI Costs
Use our full calculator to compare Gemini models against every other provider.