GPT Image 2.5 Flare & Sunburst are live on EvoLinkTry GPT Image 2.5

GPT-6.1 Terra API

Use GPT-6.1 Terra for screenshots, drawings, interface tasks and writing with a 1.05M-token context. Compare token rates and retain GPT model choices on one EvoLink API key. Tool calls use Responses.

OpenAIText GenerationAvailable
from $0.900 / 1M input tokens$1.000 official price-10%
API docs
Chat (no tools)Responses toolsReasoning: none to maxPrompt cachingText + image input
Production routeLive
Context
1.05M context · 128K max output
Best For
Visual reasoning, computer use, writing and editing
Input
Text + images
Output
Text · JSON (structured output) · tool calls

Choose GPT-6.1 Terra

One exact model ID on the OpenAI-compatible integration. Short- and long-context rates for all four token roles are listed on the card.

GPT-6.1 Terra

One versioned model ID for vision, interface and writing workloads; Responses handles tool calls, while Chat Completions accepts requests without tools.

Selected
Model ID
gpt-6.1-terra
Best for

Image and drawing questions, tool-driven interface steps and text revision with task-level acceptance checks

≤ 272K
Input
$0.900 / 1M-10%
61.2 cr / 1M$1.000official price
Cache read
$0.046 / 1M-9%
3.1 cr / 1M$0.050official price
Cache write
$1.125 / 1M-10%
76.5 cr / 1M$1.250official price
Output
$4.500 / 1M-10%
306 cr / 1M$5.000official price
> 272K
Input
$1.800 / 1M-10%
122.4 cr / 1M$2.000official price
Cache read
$0.092 / 1M-9%
6.2 cr / 1M$0.100official price
Cache write
$2.250 / 1M-10%
153 cr / 1M$2.500official 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. At more than 272K prompt tokens, input and both cache roles use 2x rates while output uses 1.5x rates for the entire request.

GPT-6.1 Terra pricing

Estimate what one GPT-6.1 Terra 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 successful tool calls.

Estimated request cost

GPT-6.1 Terra
Short-context rate
USD$0.0023
Credits0.1537

Official estimate $0.0026 · save $0.0003 (10%)

Input tokens0.0612 cr
Cache read tokens0.0007 cr
Cache write tokens0 cr
Output tokens0.0918 cr

Minimum charge: 0.01 credits per request. At more than 272K prompt tokens, input and both cache roles use 2x rates while output uses 1.5x rates for the entire request.

Budget guide

Approximate requests using the current token mix.
Add credits
$10
About 4229 requests

For quick testing

$50
About 22121 requests

For regular development

$100
About 44242 requests

For production evaluation

Server-side tool rates

Only successful server-side calls are billed per call; failed attempts have no tool fee, but tokens still apply.
  • Web search$0.010/ call0.68 cr / call

What is the GPT-6.1 Terra API?

OpenAI's fast, efficient GPT-6.1 tier for visual reasoning, precise computer use and natural writing. Supply text or images and receive text or structured results; use application tools through Responses when an answer has to become a checked action.

GPT-6.1 Terra is served on EvoLink under the model ID gpt-6.1-terra through Chat Completions · Responses, with the same API key and balance you use for every other model. It offers a 1.05M context window and up to 128K output tokens, plus image input, multi-step agents, tool use.

The GPT-6.1 Terra API lets an application use the model for visual reasoning, interface workflows and writing. With EvoLink, keep model selection, one API key and one balance in a unified integration.

GPT-6.1 Terra

GPT-6.1 Terra specs and capabilities

Numbers come from the EvoLink route configuration; capabilities are what the API exposes today.

Context window
1.05M tokens
Max output
128K tokens
Input
Text + images
Output
Text · JSON (structured output) · tool calls
Reasoning
Configurable reasoning effort
Tool use
Function calling with multi-step tool sequences
Prompt caching
Cache write + cache read rates
Long-context tier
At more than 272K prompt tokens, input and both cache roles use 2x rates while output uses 1.5x rates for the entire request.
Protocols
Chat Completions · Responses
Model ID
gpt-6.1-terra

GPT-6.1 Terra API use cases for vision, interfaces and writing

Start with image-heavy questions, verifiable interface steps and editing with clear constraints. OpenAI now publishes Terra-specific benchmarks; the results below are a starting point for your own quality and latency tests.

GPT-6.1 Terra Visual Reasoning

Supply product photos, screenshots, technical drawings or charts with a precise question, such as matching two product images or reading a dimension. Ask for the visual evidence behind each answer and check extracted values against the source; image tokens are billed as input. For invoice processing, request structured fields such as invoice number, total and currency, then check them against the original before updating accounting records.

GPT-6.1 Terra Interface Workflows on Responses

Bring your application's tools to a Responses workflow for screen reading, form filling or navigation. Keep permissions, confirmations and final-state checks under application control, and replay failures before allowing writes. Chat Completions is for requests without tools.

Writing and Editing with GPT-6.1 Terra

Use Terra for drafts, rewrites and edits that must keep meaning, terminology and length limits. Give the source text and the constraints, ask for a list of removed or changed facts, and review the result; fluent output is not proof that every fact survived the edit.

GPT-6.1 Terra Reasoning Effort

Start from medium, the default, and compare levels on the same task set. Use none for quick requests where you want temperature or top_p control; raise effort for harder visual or multi-step work only when accepted results justify the extra output tokens and latency.

Two ways to use GPT-6.1 Terra: EvoLink API or Agent

Use the EvoLink API for product backends and batch jobs, or call GPT-6.1 Terra 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 OpenAI-compatible Chat Completions (or Anthropic Messages) requests to EvoLink and control the model ID, system prompt, output budget, tools, and structured output.

  1. 1Create an EvoLink API key in the console
  2. 2Point your OpenAI or Anthropic SDK at the EvoLink base URL and select the model ID shown above
  3. 3Send one representative request and read the usage field for input, cached, and output tokens
  4. 4Set max_tokens and retries per task; keep tool-call IDs and results across turns
Option 2

Call it with an Agent

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 GPT-6.1 Terra through EvoLink, and returns the answer with token usage.

  1. 1Set EVOLINK_API_KEY in your local environment; never put it in code or a prompt
  2. 2Describe the task, the inputs to include, and the expected output format
  3. 3Ask the Agent to call GPT-6.1 Terra through EvoLink and show the request before sending
  4. 4Let the Agent report the answer, token usage, and any error body

GPT-6.1 Terra API code example and error handling

The Chat Completions example below is for text without tools. For tool calls, use the separate Responses example and its input, reasoning.effort and max_output_tokens fields.

View complete API docs
cURL
curl -X POST https://api.evolink.ai/v1/chat/completions \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6.1-terra",
    "messages": [
      { "role": "system", "content": "You are a careful editor. Preserve the original meaning and terminology." },
      { "role": "user", "content": "Tighten this release note to 120 words and list the facts you removed:\n<draft>" }
    ],
    "reasoning_effort": "medium",
    "max_tokens": 4096
  }'

# Use the exact ID gpt-6.1-terra (with a dot) on both Chat Completions and Responses.
# reasoning_effort accepts none, low, medium (default), high, xhigh, and max;
# minimal is not supported. Reasoning tokens are billed as output.
# temperature and top_p only apply with reasoning_effort none; remove logprobs and
# top_logprobs at other levels. Use Responses for function calling and built-in tools.

Invalid request or unsupported parameter

Check the chosen endpoint and its field names: messages, reasoning_effort and max_tokens for Chat Completions; input, reasoning.effort and max_output_tokens for Responses.

Authentication or balance issue

Check the Authorization bearer token and confirm the available balance in the console.

Context length exceeded

Prompt tokens exceed the GPT-6.1 Terra 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.

GPT-6.1 Terra · Responses API

The Chat Completions example below is for text without tools. For tool calls, use the separate Responses example and its input, reasoning.effort and max_output_tokens fields.

POST https://api.evolink.ai/v1/responses

curl -X POST https://api.evolink.ai/v1/responses \
  -H "Authorization: Bearer $EVOLINK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6.1-terra",
    "input": "Explain the difference between cache reads and cache writes.",
    "reasoning": { "effort": "medium" },
    "max_output_tokens": 4096
  }'
View complete API docs

Before you take GPT-6.1 Terra to production

Four checks that decide cost and stability.

01

Pick the exact GPT-6.1 Terra model ID

Use gpt-6.1-terra, with a dot, on both Chat Completions and Responses. It does not replace gpt-5.6-terra, which stays available under its own ID.

Required
02

Send tool calls to Responses

Function calling and built-in tools use the Responses API. Keep Chat Completions for requests without tools; an old Chat tool loop does not carry over by changing the model ID.

Required
03

Check reasoning effort and sampling settings

Use none, low, medium, high, xhigh or max; minimal is not supported, so start such requests at low. temperature and top_p only apply with none; remove logprobs and top_logprobs at other levels.

Required
04

Structure prompts for cache reads and budget for 272K

Put stable instructions first so cache reads replace full-price input; above 272K prompt tokens the whole request uses long-context rates.

Recommended

Four-part prompt caching

Read the four token roles separately: uncached input, cache reads, cache writes and output. The usage and pricing sections determine what is billed; a repeated prompt is not proof of a cache hit. OpenAI prices GPT-6.1 Terra cache reads at 5% of uncached input (GPT-5.6 Terra: 10%), but output, retries and review still determine whole-task cost.

Long-context threshold

At more than 272K prompt tokens, input and both cache roles use 2x rates while output uses 1.5x rates for the entire request.

GPT-6.1 Terra benchmarks and reported results

OpenAI results, not EvoLink route measurements. Compare the same benchmark version, reasoning effort and tool setup.

Benchmark / settingGPT-6.1 TerraGPT-5.6 TerraHow to read the result
Terminal-Bench 4.0 · max50.3%21.5%+28.8 percentage points
Factual error rate · low16.7%53.6%36.9 percentage points lower

Lower is better for factual errors. Benchmark scores do not establish your application’s accuracy or latency.

Source: OpenAI — Introducing GPT-6.1 Terra

Choosing a GPT-6.1 Terra API model

Choose by accepted-task cost, including retries and human review. Keep model selection on the same EvoLink integration.

GPT-6.1 Terra vs GPT-6.1 Sol

Test Terra on screenshots, interface tasks and constrained edits. Keep Sol in the evaluation for complex coding and long agent runs; a cheaper token rate only helps when the accepted result also costs less.

GPT-6.1 Terra vs GPT-6 Luna

Keep Luna as a candidate for focused, high-volume classification and extraction. Test Terra where visual detail or interface state makes the task harder, and compare completed-task cost at fixed acceptance criteria.

GPT-6.1 Terra vs GPT-6 Astra

Include Astra when the work needs deeper reasoning or difficult professional judgment. Compare it with Terra on the same inputs instead of treating a computer-use benchmark as proof of equal capability on every task.

GPT-6.1 Terra pricing and GPT model choices

EvoLink

Compare EvoLink input/output rates and workload fit for 6.1 Terra, 6.1 Sol and the previous Terra. See the Models card for cache and long-context rates, and Pricing for request estimates.

GPT-6.1 Terra
EvoLink price / 1M (input / output)$0.9 / $4.5
Context window1.05M
Prompt cachingCache read + cache write
Best forImage and drawing questions, tool-driven interface steps and text revision with task-level acceptance checks
GPT-6.1 Sol
EvoLink price / 1M (input / output)$1.8 / $9
Context window1.05M
Prompt cachingCache read + cache write
Best forHarder multi-file coding and long agent runs; no reasoning effort none
GPT-5.6 Terra
EvoLink price / 1M (input / output)$1.8 / $10.8
Context window1.05M
Prompt cachingCache read + cache write
Best forPrevious Terra release at higher OpenAI list rates, with the same context and reasoning levels

Base-context prices reflect your account's current pricing; signed-out visitors see Default rates. Long-context multipliers are listed in the Pricing section.

GPT Model Family

Same API key and balance — switch tiers without changing your integration.

Compare all GPT models
GPT-6.1 Sol

GPT-6.1 Sol

Near-Astra performance for complex work at a lower cost

View model
GPT-6 Astra

GPT-6 Astra

OpenAI's most capable model for the hardest end-to-end work

View model
GPT-6 Sol

GPT-6 Sol

GPT-6 model built for complex coding and agentic workflows

View model

Other text models on EvoLink besides GPT-6.1 Terra

Claude Opus 5.5

Claude Opus 5.5

Anthropic's newest Opus model for the hardest long-horizon coding and agent work, with a 1M context window and thinking always on.

View model
Claude Sonnet 5.5

Claude Sonnet 5.5

Anthropic's newest Sonnet model for everyday coding, tool-using agents, and high-volume production traffic, with a 1M context window.

View model
Grok 4.7

Grok 4.7

xAI reasoning and tool-use model with a 500K context window and metered server-side tools.

View model
Gemini 3.8 Flash

Gemini 3.8 Flash

Google's most capable Flash model, with gains on software engineering and agentic tasks and a 1M context window.

View model

GPT-6.1 Terra guides and related reading

GPT-6.1 Terra: background and release news

GPT-6.1 Terra: background and release news

Written before the October 26 release; use this page for the current model ID, parameters and rates

Read guide
GPT-5.6 Sol vs Terra vs Luna routing guide

GPT-5.6 Sol vs Terra vs Luna routing guide

How production teams split traffic across the Sol, Terra and Luna tiers

Read guide
GPT-6 Astra API guide

GPT-6 Astra API guide

GPT-6 family integration reference; use this page for Terra-specific request settings

Read guide

GPT-6.1 Terra API - FAQ

Which model ID should I use?

Use gpt-6.1-terra, with a dot. The same ID works on Chat Completions and Responses. gpt-5.6-terra is a separate model and stays available under its own ID.

How much does GPT-6.1 Terra cost on EvoLink?

For prompts up to 272K tokens, the public Default reference rates are $0.900 input and $4.500 output per 1M tokens; cache reads are $0.046 and 30-minute cache writes $1.125. Signed-in account rates may differ: use the Models card and calculator for your current prices. OpenAI Standard input/output list rates are $1.000 / $5.000. Longer prompts use the separate multipliers shown on this page.

How is cached input billed?

Cache reads and cache writes are separate token segments, and neither is billed again as ordinary input. OpenAI's cache read price for GPT-6.1 Terra is 95% below uncached input; for GPT-5.6 Terra it was 90% below.

When do long-context rates apply?

When total prompt tokens exceed 272,000, the whole request uses the long-context multipliers: 2x for input and both cache roles, 1.5x for output.

Which reasoning effort levels are supported?

none, low, medium (the default), high, xhigh and max. minimal is not supported; requests that still send it should start at low. Reasoning tokens are billed as output.

Can I use function calling on Chat Completions?

Use the Responses API for function calling and built-in tools with the same gpt-6.1-terra ID. Chat Completions is for requests without tools, including requests at reasoning effort none.

Are temperature, top_p, and logprobs supported?

Only with reasoning effort none. At any other level, including the default medium, temperature and top_p are not applied; remove logprobs and top_logprobs, and drop message.output_text.logprobs from the Responses include list.

What are the context window and maximum output?

The context window is 1,050,000 tokens, with a maximum output of 128,000 tokens. Leave room for the response when sizing your prompt. OpenAI lists June 1, 2026 as the knowledge cutoff. The model takes text and images and directly outputs text.

Should I move from GPT-5.6 Terra to GPT-6.1 Terra?

OpenAI Standard list rates are lower for every token role: input and cache writes are half, output is about 58% lower and cache reads are 75% lower. Context, maximum output and reasoning levels are unchanged. Move tool calls to Responses, then compare accepted-task cost and latency before shifting traffic.

GPT-6.1 Terra or GPT-6.1 Sol: which one should I use?

Compare accepted results and total cost on your own tasks. GPT-6.1 Terra's OpenAI Standard input/output rates are half of Sol's, and Terra adds reasoning effort none. Keep Sol for the hardest coding and long agent runs where its measured quality justifies the cost; retain both model IDs.

Is GPT-6.1 Terra more than 3× faster than Sol and Astra?

OpenAI reports over 3× faster task completion than GPT-6.1 Sol and GPT-6 Astra in its Agents’ Last Exam evaluation. This is a benchmark-specific claim, not an EvoLink latency guarantee or a claim about every API request. Measure time to first token and end-to-end completion on your workload.