GPT API Family
Use one EvoLink API to access GPT models. Compare GPT-6.1 Sol, GPT-6 Astra, Sol, and Luna, GPT-5.6, GPT-5.5, GPT-5.4, GPT-5.2, and GPT-5.1 on API pricing, context window, reasoning fit, and capabilities — then pick the right model for your workload.
9 models
Covers different strengths and cost tiers
Unified API access
OpenAI compatible, fast onboarding
Choose by workflow
Match model to task before integrating
Compare the GPT models
Choose based on your primary constraint: reasoning depth, context length, or cost.
| Model | Best for | Input / MTok | Output / MTok | Context | Cached input |
|---|---|---|---|---|---|
GPT-6.1 Sol Coding agents | Multi-file changes, document questions and tool-driven agents with the exact ID gpt-6.1-sol. Use low to xhigh reasoning on either endpoint, max on Responses, and Responses for tool calling. | $1.80 | $9.00 | 1.05M | $0.092 |
GPT-6 Astra Flagship | OpenAI's most capable model — frontier reasoning, agentic coding, computer use, and research on the exact ID gpt-6-astra. | $9.00 | $45.00 | 1.05M | $0.90 |
GPT-6 Sol Coding & Agents | GPT-6 model built for complex coding and agentic workflows, with reasoning effort from none to max on the exact ID gpt-6-sol. | $1.80 | $9.00 | 1.05M | $0.181 |
GPT-6 Luna High Volume | The most efficient GPT-6 model for focused, high-volume tasks such as classification, extraction, and routing, on the exact ID gpt-6-luna. | $0.092 | $0.45 | 1.05M | $0.0089 |
GPT-5.6 Live | Three-tier GPT-5.6 family — Sol for frontier reasoning, Terra for balanced production, Luna for high-volume cost control. | from $0.181 | from $1.081 | 1.05M | from $0.02 |
GPT-5.5 Previous Flagship | Advanced reasoning, demanding workflows, and newer GPT-5.5 API evaluation. | $4.50 | $27.00 | 1M | $0.45 |
GPT-5.4 Previous Flagship | Complex reasoning, 1M+ context analysis, computer use, and agent orchestration. | $2.25 | $13.50 | 1.05M | $0.225 |
GPT-5.2 Best Value | Production coding, reasoning, and multi-turn conversations at 400K context. | $1.483 | $11.859 | 400K | $0.149 |
GPT-5.1 Budget | High-volume tasks where cost matters most: summarization, classification, and generation. | $1.059 | $8.471 | 400K | $0.111 |
How to decide which GPT model to use
Follow these 4 rules to narrow down your choice.
Start with task complexity
Frontier reasoning, agentic coding, computer use, and high-accuracy output — start with GPT-6 Astra, or evaluate GPT-6.1 Sol for multi-file coding and tool-driven agents against your own acceptance and task-cost criteria.
Then check context length
Long documents, codebases, logs, research papers, multi-turn history — compare GPT-5.4.
Then check call frequency
Support, summarization, classification, tagging, batch text processing — compare GPT-6 Luna, GPT-5.2, or GPT-5.1.
Finally, consider whether to fix a model
If the same workflow mixes light and heavy tasks, consider EvoLink Smart Router instead of hardcoding one model for every step.
Smart Router →If you already know your task type, find the recommended starting point in the table below.
Choose a GPT model by workflow: reasoning, coding, summarization, and high-volume tasks
Match your primary task to the right GPT model.
| Your task | Recommended start | Good fit if... | Watch out for |
|---|---|---|---|
| Frontier reasoning, agentic coding, computer use | GPT-6 Astra | You need OpenAI's most capable model for the hardest end-to-end reasoning, coding, computer-use, or research work, and output quality justifies $9.00 / $45.00 per 1M tokens | Test GPT-6.1 Sol on the same accepted-task, latency and cost criteria before switching |
| Multi-file coding and tool-driven agents | GPT-6.1 Sol | You need multi-file changes or multi-step agents at displayed EvoLink input/output rates of $1.80 / $9.00 per 1M tokens for prompts up to 272K. Evaluate accepted-task cost on your own fixed jobs; token rates do not establish whole-task savings or Astra-equivalent quality. | Tool calling needs the Responses API; reasoning effort starts at low (there is no none), and temperature, top_p, and logprobs are rejected |
| Complex coding and agentic workflows | GPT-6 Sol | You run coding agents, multi-file changes, or tool-heavy workflows and want the GPT-6 generation at $1.80 / $9.00 per 1M tokens | Use the Responses API for built-in tools; on Chat Completions, function calling requires reasoning_effort none |
| High-volume classification, extraction, routing | GPT-6 Luna | Calls are short, focused, and frequent, and latency and unit cost matter most | Start at reasoning_effort none and raise it only where evaluation shows a quality gain |
| Typed answers: a probability, a fixed option, or a score | GPT-6 Luna Decisions | You need the probability that a condition is true, one option from a fixed set, or a rubric score for text or images, not generated text | Runs only on POST /v1/decisions: no streaming, system prompts, or tools, and only input tokens are billed |
| Frontier reasoning, agentic coding, high-value analysis | GPT-5.6 Sol / Terra | You need the strongest reasoning path, coding agents, tool use, or output quality that justifies higher model cost | Use Sol for the hardest tasks; test Terra when you need a stronger cost-latency balance |
| Complex reasoning and coding | GPT-5.5 | You need higher accuracy, multi-step reasoning, code generation, tool use, or complex problem decomposition | Good fallback benchmark if GPT-5.6 cost or latency is too high |
| Long document or codebase analysis | GPT-5.4 | Your input is long — contracts, papers, logs, codebases, or multi-turn context | Watch input token cost — estimate context size first |
| Everyday Q&A, summarization, classification | GPT-5.2 | You need stable results while keeping cost under control | Good default starting point for most production tasks |
| High-frequency lightweight tasks | GPT-5.1 | Tasks are simple, call volume is high, cost sensitivity is primary | Not suited for complex reasoning or high-value outputs |
| Mixed-complexity text tasks | EvoLink Smart Router | Same workflow has both simple and complex tasks | Best when you don't want to maintain manual model routing logic |
GPT API workflows: agents, chat, summarization, and content processing
See how GPT models fit into real products, agents, and content processing pipelines.
Reasoning and coding
For code generation, bug fixing, test case generation, complex logic analysis, and tool-calling agents. If the output directly affects product quality or development efficiency, start testing with GPT-5.5. If context is especially long, compare GPT-5.4.
High-volume chat and support
For support bots, in-app assistants, knowledge base Q&A, and high-frequency multi-turn conversations. If per-request value is low but call volume is high, test with GPT-5.2 first, then stress-test cost with GPT-5.1.
Summarization and classification
For long-text summaries, tag classification, structured extraction, review categorization, and batch content processing. These tasks usually don't need the strongest model — GPT-5.2 often strikes a better balance between quality and cost.
Agent routing and mixed text tasks
For workflows where simple classification, retrieval, reasoning, and generation coexist in the same pipeline. If you don't want to hardcode a model for every step, use EvoLink Smart Router to handle routing at the API layer via evolink/auto.
Explore each GPT model
Use this page to compare, then visit individual model pages for pricing details, playground access, and integration guides.
GPT-6.1 Sol
Coding agents
- Context
- 1.05M
- Pricing
- $1.80 / $9.00
GPT-6 Astra
Flagship
- Context
- 1.05M
- Pricing
- $9.00 / $45.00
GPT-6 Sol
Coding & Agents
- Context
- 1.05M
- Pricing
- $1.80 / $9.00
GPT-6 Luna
High Volume
- Context
- 1.05M
- Pricing
- $0.092 / $0.45
GPT-5.6
Live
- Context
- 1.05M
- Pricing
- from $0.181 / from $1.081
GPT-5.5
Previous Flagship
- Context
- 1M
- Pricing
- $4.50 / $27.00
GPT-5.4
Previous Flagship
- Context
- 1.05M
- Pricing
- $2.25 / $13.50
GPT-5.2
Best Value
- Context
- 400K
- Pricing
- $1.483 / $11.859
GPT-5.1
Budget
- Context
- 400K
- Pricing
- $1.059 / $8.471
Access GPT models through one EvoLink API
Use one EvoLink API key to choose among GPT-6.1 Sol, GPT-6 Astra, Sol and Luna, GPT-5.6, GPT-5.5, GPT-5.4, GPT-5.2 and GPT-5.1. Match the endpoint and parameters to each model before switching; GPT-6.1 Sol tool calls use Responses.
Switch model="gpt-5.5" to model="gpt-5.2" without rebuilding your integration.How to think about GPT API cost: complex reasoning, long context, and high-frequency tasks
Complex reasoning amplifies output cost
Complex reasoning, code generation, and multi-turn tool use tend to produce longer outputs and work best with higher-capability models. If task value is high, GPT-5.5 quality may matter more than cost. For simple tasks, don't default to the most capable model.
Long context amplifies input cost
Document analysis, codebase understanding, log processing, and research summarization bring large input token volumes. If the bottleneck is input length rather than reasoning depth, GPT-5.4 may be more appropriate.
High-frequency tasks need low unit cost
Support, summarization, classification, and tagging at high volume should prioritize unit cost. Test with GPT-5.2 for quality first, then try GPT-5.1 to see if you can reduce cost further.
Pricing summary
Token-based billing with separate input, output, and cache rates. Each model page lists current EvoLink pricing.
GPT-6.1 Sol
$1.80 input
$9.00 output
Context: 1.05M
EvoLink GPT-6.1 Sol · Input $1.80 · Output $9.00 / 1M tokens · 1.05M context, 128K max output, cache reads and writes.
GPT-6 Astra
$9.00 input
$45.00 output
Context: 1.05M
EvoLink GPT-6 Astra · Input $9.00 · Output $45.00 / 1M tokens · 1.05M context, 128K max output, cache reads and writes.
GPT-6 Sol
$1.80 input
$9.00 output
Context: 1.05M
EvoLink GPT-6 Sol · Input $1.80 · Output $9.00 / 1M tokens · 1.05M context, 128K max output, cache reads and writes.
GPT-6 Luna
$0.092 input
$0.45 output
Context: 1.05M
EvoLink GPT-6 Luna · Input $0.092 · Output $0.45 / 1M tokens · 1.05M context, 128K max output, cache reads and writes.
GPT-5.6
from $0.181 input
from $1.081 output
Context: 1.05M
GPT-5.6 family (Sol, Terra, Luna) is now GA and live on EvoLink routing, with 1.05M context, 128K max output, and cache reads and writes. OpenAI list pricing starts at $0.20/$1.20 per 1M tokens for Luna.
GPT-5.5
$4.50 input
$27.00 output
Context: 1M
EvoLink GPT-5.5 · Input $4.50 · Output $27.00 / 1M tokens.
GPT-5.4
$2.25 input
$13.50 output
Context: 1.05M
EvoLink GPT-5.4 · Input $2.25 · Output $13.50 / 1M tokens.
GPT-5.2
$1.483 input
$11.859 output
Context: 400K
Best value for production reasoning and coding workloads at 400K context.
GPT-5.1
$1.059 input
$8.471 output
Context: 400K
Budget tier for high-volume tasks where cost matters most.
Related GPT guides
Use the family page to compare models, then visit guides for pricing details, comparisons, and integration.
GPT-5.6 Sol vs Terra vs Luna Routing Guide
Choose the right GPT-5.6 tier for production workloads: Sol for frontier reasoning, Terra for balanced traffic, Luna for high-volume cost control.
GPT-5.4 vs Claude Opus 4.6 vs Gemini 3.1 Pro
Three-way flagship comparison with pricing tables, benchmarks, and decision framework.
GPT-5.4 Pricing: Tiers, Caching & EvoLink Rates
GPT-5.4 pricing details covering the >272K input tier, cached-input rules, and how to compare current EvoLink rates.
GPT-5.2 vs Gemini 3 Pro Comparison
Head-to-head comparison of GPT-5.2 and Gemini 3 Pro on pricing, benchmarks, and use cases.
Best LLM for Coding Agents
Compare Claude, GPT, DeepSeek, Qwen Coder, and Gemini for coding agent workloads — API cost, tool-call reliability, and fallback planning.
GPT-6 vs GPT-5.6: Upgrade or Stay?
When the GPT-6 Astra upgrade pays off and when GPT-5.6 Sol, Terra, or Luna remains the better route for cost and latency.
GPT API Family FAQ
Everything you need to know about the product and billing.