Use Gemini 4 Hydrogen for text and image tasks with low, medium, high or xhigh thinking. Connect through EvoLink’s unified API for structured extraction and tool calling, using OpenAI-compatible Chat Completions or the Gemini native API with one API key.
Google·Text Generation·Available
From $1.350 / 1M input tokens$1.500 official price-10%
Thinking levelsImage inputFunction callingGoogle SearchGemini native API
Production routeThis rate reflects platform-side availability — only confirmed server errors (HTTP 500 / empty response) count as failures. User-side issues (content moderation, invalid params, cancellation) plus rate limits, timeouts and auth errors are excluded. Before real traffic arrives, empty buckets may display as available.Live
Copy the Gemini 4 Hydrogen model ID, check current rates, and connect with your existing EvoLink API key.
Gemini 4 Hydrogen
Lower-priced Gemini 4 model
Selected
Model ID
gemini-4-hydrogen
Best for
Run high-volume chat, extraction and tool-calling tasks. Measure thinking tokens and search queries, then compare accepted results and total task cost with Gemini 3.8 Flash before switching production traffic.
Input
$1.350 / 1M-10%
91.8 cr / 1M$1.500official price
Cache read
$0.068 / 1M-10%
4.6 cr / 1M$0.075official price
Output
$6.750 / 1M-10%
459 cr / 1M$7.500official price
All rates are per 1M tokens, shown in USD and credits, and reflect your account's current pricing.
Gemini 4 Hydrogen cost estimate
Estimate what one Gemini 4 Hydrogen request costs before you integrate. The calculator uses your account's current rates, with official pricing as a reference.
Request calculator
Enter the token mix for one request and the number of Google Search queries.
Estimated request cost
Gemini 4 Hydrogen
USD$0.0034
Credits0.2305
Official estimate $0.0038 · save $0.0004 (10%)
Input tokens0.0918 cr
Cache read tokens0.001 cr
Output tokens0.1377 cr
Google Search0 cr
Minimum charge: 0.01 credits per request.
Budget guide
Approximate requests using the current token mix and search queries.
Gemini 4 Hydrogen bills each nonempty search query, even if no sources are returned. One request may run multiple queries. Search fees are added separately after the token minimum charge.
Google Search$0.014/ query·0.952 cr / query
What is Gemini 4 Hydrogen?
Gemini 4 Hydrogen is a Google Gemini 4 model served on EvoLink as gemini-4-hydrogen. It accepts text and image inputs and returns text, structured JSON or function calls. Thinking levels, Google Search and automatic prefix caching let you control how a task runs. Use the workflows below to assess image extraction, support answers and tool use; token rates alone do not establish which model delivers the lowest task cost.
Gemini 4 Hydrogen specs and capabilities
Numbers come from the EvoLink route configuration; capabilities are what the API exposes today.
Context window
1.05M tokens
Max output
1.05M tokens
Input
Text + images + video + files
Output
Text · JSON (structured output) · tool calls
Reasoning
Configurable reasoning effort
Tool use
Function calling with multi-step tool sequences
Prompt caching
Automatic cache reads at a lower rate
Server-side tools
Google Search, billed per search query
Protocols
Chat Completions · Gemini native API
Model ID
gemini-4-hydrogen
Gemini 4 Hydrogen use cases
Choose a workflow by its input, required output and failure conditions. Keep application-side validation and human review where errors would affect a user or a business record.
Gemini 4 Hydrogen for image-to-JSON extraction
Send an invoice screenshot or product image with a fixed field schema. Ask for the values visible in the image and null for missing or unreadable fields. Validate types, amounts and required fields before writing a record; route uncertain extractions for review. Track field accuracy and cost per accepted record, including retries.
Gemini 4 Hydrogen for knowledge-based support
Place shared support instructions and policy documents before the customer’s changing question and conversation. Require answers to stay within that evidence and escalate when it is missing or conflicting. Repeated prefixes may benefit from caching; measure cache reads as well as answer acceptance, response time and escalation rate.
Gemini 4 Hydrogen for tool-calling agents
For an order-status task, let the model propose a lookup using your order ID schema; your application validates and runs the function, then returns its result for the final answer. Preserve thought signatures with call history. Allow Google Search only when the task needs web evidence, and cap queries, retries and tool rounds. A valid tool call is not proof the task completed.
Gemini 4 Hydrogen image-to-JSON example
Illustrative task and acceptance contract, not a recorded model response or benchmark. Use a clear image, a field schema and a labelled reference to evaluate the result.
Input and requested fields
Suppose an invoice image clearly shows ACME Corp, USD 1,240.50 and a due date of 2026-12-01. Request vendor, currency, total and due_date as structured JSON. Instruct the model to use null for absent or unreadable values and to avoid filling gaps from general knowledge.
Acceptance before saving
The expected field values for this illustrative input are vendor = ACME Corp, currency = USD, total = 1240.50 and due_date = 2026-12-01. Your application checks the JSON schema and compares the values with the source. A schema-valid object with the wrong amount fails; an unreadable amount goes to review rather than an automatic retry loop.
Two ways to use Gemini 4 Hydrogen: EvoLink API or Agent
Use the EvoLink API for product backends and batch jobs, or call Gemini 4 Hydrogen from Codex, Claude, or Gemini for coding and analysis workflows. Both paths share the same EvoLink API key, balance, model ID, and request history.
Option 1
Integrate with the EvoLink API
Best for: product backends, batch jobs, automated pipelines
Send Gemini native generateContent requests to EvoLink and control the model ID, system instruction, output budget, tools, and structured output.
1Create an EvoLink API key in the console
2Call the generateContent endpoint on the EvoLink base URL with the model ID shown above in the request path
3Send one representative request and read usageMetadata for input, cached, and output tokens
4Set maxOutputTokens and retries per task; keep function-call parts and their results across turns
Best for: coding, review, and analysis tasks in Codex, Claude, and Gemini
Give the Agent the task, the inputs to include, and the acceptance criteria. It assembles the request, calls Gemini 4 Hydrogen through EvoLink, and returns the answer with token usage.
1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
2Describe the task, the inputs to include, and the expected output format
3Ask the Agent to call Gemini 4 Hydrogen through EvoLink and show the request before sending
4Let the Agent report the answer, token usage, and any error body
Gemini 4 Hydrogen API code example and error handling
This example shows the shortest runnable request: a Gemini native generateContent call with a system instruction, a user message, and an output budget. Open the API tab for the complete parameter and response reference.
# Gemini native API
curl -X POST "https://direct.evolink.ai/v1beta/models/gemini-4-hydrogen:generateContent" \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"systemInstruction": { "parts": [{ "text": "Extract the fields and answer in JSON only." }] },
"contents": [
{ "role": "user", "parts": [{ "text": "Invoice: ACME Corp, total $1,240.50, due 2026-12-01" }] }
],
"generationConfig": {
"maxOutputTokens": 2048,
"responseMimeType": "application/json",
"thinkingConfig": { "thinkingLevel": "low" }
}
}'
# thinkingLevel accepts low, medium, high (default), or xhigh. Gemini 4 Hydrogen thinks at
# every level, and thinking tokens are billed at the output rate.
# For streaming, call streamGenerateContent?alt=sse with the same body.
# The response includes candidates[0].content.parts and usageMetadata
# (prompt, thinking, and output token counts).
Invalid request or unsupported parameter
Check the model ID, the contents array, and parameter ranges against the API reference; remove fields this route does not support.
Authentication or balance issue
Check the Authorization bearer token and confirm the available balance in the console.
Context length exceeded
Prompt tokens exceed the Gemini 4 Hydrogen context window. Trim or retrieve only the relevant evidence and reuse cached prefixes.
Rate limited (429)
Back off and retry with jitter; batch or queue requests instead of sending parallel bursts.
Content or tool call rejected
Review sensitive content, malformed tool-call arguments, and JSON schema mismatches before retrying.
Gemini 4 Hydrogen API integration checks
Use your EvoLink API key with the protocol your application already supports. Check these four items before routing production tasks to Hydrogen.
01
Set the model ID
For Chat Completions, set model to "gemini-4-hydrogen" in the request body. For the native API, use that ID in /v1beta/models/{model}:generateContent. The two protocols have different request shapes.
Required
02
Choose a thinking level
Set reasoning_effort in Chat Completions, or thinkingLevel in the native thinkingConfig, to low, medium, high or xhigh; omitting it uses high. Start with a level that meets your task’s acceptance checks; test another level when errors remain. Higher is not a quality guarantee, and low can still incur thinking costs. Keep the chosen level in your usage logs.
Cost control
03
Budget output and tools
Set maxOutputTokens (max_tokens in Chat Completions) for each request; output includes thinking tokens. Google Search is billed per query on top of tokens, and one request can run several queries. Cap tool rounds and retries for the whole task.
Cost control
04
Compare before you switch
Keep the current model when it already meets quality and cost targets. Trial Hydrogen on a small, representative slice, including difficult and failed cases. Change the route only if accepted outcomes justify the total spend and response time. Keep a tested fallback and revert when validation or tool completion fails.
Evaluate
Gemini 4 Hydrogen thinking tokens and cost
Thinking levels control a request; they do not set a fixed token allowance or guarantee a quality or cost improvement. Check schema and field accuracy for extraction, evidence use for support, and tool completion for agents. Keep the least expensive setting that passes those checks on your sample. Thinking tokens are billed at the output rate ($6.750 per 1M tokens); the usage example below explains how to avoid counting them twice.
One 1,048,576-token window; up to 65,536 tokens per streaming request
Input, thinking, and output share one 1,048,576-token context window; requests with more input are rejected. A single streaming request returns at most 65,536 thinking and output tokens. To go further on the native API, set maxOutputTokens above that limit: a response that ends with the finish reason CONTINUATION carries a continuationToken, and resending the same request with that token continues the output until it ends with STOP or reaches maxOutputTokens. Chat Completions cannot continue; a response cut at the limit ends with finish_reason "length".
Reading Gemini 4 Hydrogen usage and task cost
Token counts below are illustrative. In the existing calculator, enter uncached input, cache reads and billable output separately. If reported prompt usage includes cache hits, subtract that cached portion before filling input. Add search queries separately.
Native API: combine visible output and thinking
If usageMetadata reports 300 visible output tokens and 200 thinking tokens, those represent 500 tokens billed at the output rate. Read input and cache usage separately. This example explains the billing categories; it does not predict how much thinking Hydrogen will use.
Chat Completions: avoid counting thinking twice
If completion_tokens is 500 and its reasoning detail reports 200 tokens, the 200 are already part of the 500. Do not enter 700 as output. For a multi-request task, include every attempt’s token and tool charges, then divide total spend by the number of accepted tasks to compare routes.
Gemini 4 Hydrogen vs Gemini 4 Argon and Gemini 3.8 Flash
EvoLink
Use the current rates below as the pricing baseline. A newer model is not automatically cheaper: Gemini 3.8 Flash can have lower token rates than Hydrogen. Token rates do not measure answer quality, retry volume or task completion. The selection notes below explain when to keep your current route and when to evaluate another one.
Best forAn existing Flash baseline for coding, reasoning and agents; retain it when quality, response time and total cost already meet your targets.
Choosing a Gemini route for your workload
These are routing decisions, not a benchmark ranking. Use the live comparison table above and the requirements of the task you already run.
Keep Gemini 3.8 Flash when it already works
Retain your existing route when it passes acceptance checks and meets the response-time and cost targets. Lower current Flash token rates can outweigh moving to a newer model. Introduce Hydrogen only if measured task outcomes justify the change; a new generation is not a reason to migrate by itself.
Evaluate Hydrogen for a text-and-image workflow
Trial image-to-JSON extraction, evidence-based support or function calling with the thinking setting you intend to deploy. Compare field errors, escalations, completed tool tasks, retries and billed usage. Its EvoLink API key and protocol options can fit an existing client, but model controls and tool history still need validation.
Evaluate Argon when Hydrogen misses a requirement
Compare Argon on the same difficult cases when Hydrogen misses your acceptance checks. Check capabilities and protocol limits on each model page before moving traffic; higher token rates do not guarantee a better result. For long-output workflows, verify continuation behavior and the shared context budget for the protocol you use.
Gemini Model Family
Same API key and balance — switch tiers without changing your integration.
Can I access the Gemini 4 Hydrogen API through EvoLink?
Use your EvoLink API key with "gemini-4-hydrogen" on direct.evolink.ai. This route supports OpenAI-compatible Chat Completions and the Gemini native API. Access through Google AI Studio, the Gemini app or other providers is separate.
What model ID should I use for the Gemini 4 Hydrogen API?
Use "gemini-4-hydrogen". For Chat Completions, set it as model in the request body; for the native API, use it in /v1beta/models/{model}:generateContent. The two protocols have different request shapes.
How much does the Gemini 4 Hydrogen API cost on EvoLink?
$1.350 per 1M input tokens, $6.750 per 1M output tokens and $0.068 per 1M cached input tokens. Thinking tokens are billed at the output rate. Google Search queries incur an additional charge. Displayed discounts use comparison rates of $1.500 input, $7.500 output and $0.075 cached input per 1M tokens. Use the pricing calculator for an estimate, then reconcile usage and retries to measure the cost per accepted task.
What is Gemini 4 Hydrogen useful for?
Evaluate image-to-JSON extraction, support answers grounded in your documents, and function-calling workflows. The Readme shows an illustrative invoice task and acceptance checks. Review uncertain fields and incomplete tool tasks rather than assuming structured output is correct.
How do thinking levels affect Gemini 4 Hydrogen pricing?
Set reasoning_effort or native thinkingLevel to low, medium, high or xhigh; Hydrogen uses high when you omit it. xhigh thinks the longest and adds the most latency and output tokens. All levels can incur thinking tokens billed as output; higher does not guarantee a better answer. Numeric thinking budgets are not supported. The Readme explains how native and Chat Completions usage count thinking.
Which inputs, outputs and tools are supported on EvoLink?
Text, image, audio, video, and PDF input, structured JSON output, function calling, and Google Search work on both protocols. Word, PowerPoint, and Excel files work on the Gemini native API as inline data or a file URL; on Chat Completions, send them as base64 file data, because Office file URLs are not accepted there. When you send function-call history back, include the thought signatures from the previous response.
How does Gemini 4 Hydrogen prompt caching work?
Repeated prompt prefixes can be cached automatically and cache reads are billed at the cached-input rate. Place stable instructions and shared documents before changing user content. Check cached-input usage instead of assuming every request gets a cache hit. Explicit cache creation is not supported on this route.
What is Gemini 4 Hydrogen's context window and output limit?
Input, thinking, and output share one 1,048,576-token window, and requests with more input are rejected. A single streaming request returns at most 65,536 thinking and output tokens. On the native API, longer outputs continue across requests: when a response ends with the finish reason CONTINUATION, resend the same request with its continuationToken until it ends with STOP. maxOutputTokens is the total budget across those requests.
Does Gemini 4 Hydrogen support OpenAI-compatible requests and streaming?
Yes. Use OpenAI-compatible Chat Completions or the Gemini native API with the same EvoLink API key and model ID on direct.evolink.ai. Both support streaming; the native API uses streamGenerateContent with alt=sse. Read usage from the final chunk to reconcile billed tokens. Request bodies and tool history must follow the selected protocol.
How should I choose between Gemini 4 Hydrogen and Gemini 4 Argon?
Hydrogen has lower token rates than Argon on EvoLink, but that does not prove a lower cost per completed task. Keep the route that meets your requirements after accounting for retries and tool use. Switching IDs still requires checking controls and history; a distilled architecture is not established.
Is Gemini 4 Hydrogen the same as Gemini 4 Flash?
Community discussions may use Gemini 4 Flash as a label for Hydrogen. That label is not an API alias on this page and does not establish an official name or architecture. For the EvoLink route documented here, use "gemini-4-hydrogen" rather than guessing a Flash model ID.
