Gemini Omni coming soonLearn more
Claude collection

Claude API Family

Use one EvoLink API to access every Claude model. Compare Opus, Sonnet, and Haiku on pricing, context, coding benchmarks, and capabilities — then pick the right model for your workload.

Compare the Claude models

Choose based on your primary constraint: quality, cost, context length, or output size.

ModelBest forInput / MTokOutput / MTokContextMax outputSWE-bench
Claude Opus 4.7

Current Flagship

Complex coding, agent orchestration, extended thinking, and research synthesis.$5.00$25.001M128K
Claude Opus 4.6

Previous Opus

Previous-generation Opus route for teams migrating premium coding and agent workloads to Opus 4.7.$5.00$25.001M128K80.8%
Claude Sonnet 4.6

Best Balance

Everyday coding, agentic workflows, and extended thinking at balanced cost.$3.00$15.001M128K
High-volume agents, sub-agents, latency-sensitive chat, and free-tier products.$1.00$5.00200K73.3%
Claude Sonnet 4.5

Previous Sonnet

Previous-generation Sonnet route for existing coding and agent workflows.$3.00$15.00200K (1M beta)
Previous-gen flagship with pinned versioned ID for stable production.$5.00$25.00200K64K
Multi-file refactoring, agentic coding, and deep research analysis.$15.00$75.00200K74.5%
Production workloads with 1M context beta for large documents.$3.00$15.00200K (1M beta)

How to decide which Claude model to use

Follow these 4 rules to narrow down your choice.

1

Start with task complexity

Complex reasoning, agentic coding, extended thinking, multi-step analysis — start with Opus.

2

Then check cost sensitivity

If you need strong results at lower cost, Sonnet delivers 80%+ of Opus quality at 40% less. Haiku is 5x cheaper for simple tasks.

3

Then check context and output needs

Opus 4.7, Opus 4.6, and Sonnet 4.6 support 1M context. Older Claude routes remain at 200K, with Sonnet 4.5 offering 1M beta.

4

Finally, consider routing

If tasks vary in complexity within the same workflow, consider EvoLink Smart Router instead of hardcoding one model.

Smart Router →

If you already know your task type, find the recommended starting point in the table below.

Choose a Claude model by workflow: coding, reasoning, support, and agents

Match your primary task to the right Claude model.

Your taskRecommended startGood fit if...Watch out for
Coding agentsClaude SonnetYou need balanced coding, reasoning, and cost for agentic workflowsStart with Sonnet 4.6 for new agentic coding workflows; keep Sonnet 4.5 for existing routes that depend on it
Deep reasoning and analysisClaude OpusTasks require extended thinking, multi-step reasoning, or highest quality outputHigher cost — reserve for high-value tasks
High-volume assistants and sub-agentsClaude HaikuTasks are simple, call volume is high, latency mattersNot suited for complex reasoning
Long-context document analysisClaude Sonnet 4.5You need 1M context beta for large documents or codebases1M context is in beta
Mixed text workflowsEvoLink Smart RouterSame workflow has both simple and complex tasksBest when you don't want to maintain manual routing logic

Claude API workflows

Choose the right Claude model for real production workflows.

Coding agents

Use Sonnet or Opus for code generation, review, and debugging in agentic coding workflows.

Long-context document analysis

Process large documents, contracts, and research papers with up to 1M context on the latest Opus and Sonnet routes, while older Claude routes remain at 200K.

Customer support assistants

Use Haiku for high-volume, low-latency support agents. Upgrade to Sonnet for complex reasoning when needed.

Agent routing and mixed text tasks

For supported text and agent workflows, EvoLink Smart Router can help route tasks through evolink/auto when you do not want to hardcode a single model.

View Claude model details

Pick a model, then visit its detail page for full capabilities, pricing, and integration.

Access Claude models through one EvoLink API

All Claude models are available through a single EvoLink API key and OpenAI-compatible endpoint. Switch between Opus, Sonnet, and Haiku by changing the model parameter — no separate accounts or keys needed.

One API key for all Claude models
OpenAI-compatible endpoint
Switch models by changing the model parameter
Unified billing and usage visibility

How to think about Claude API cost: reasoning depth, volume, and context

Deep reasoning amplifies output cost

Extended thinking, complex agents, and multi-step analysis produce longer outputs. If task value is high, Opus quality may matter more than cost. For simpler tasks, Sonnet or Haiku will be more cost-efficient.

High volume favors Haiku

Support bots, sub-agents, and free-tier products with high call volume should start with Haiku at $1/$5 per MTok — 5x cheaper than Opus, 3x cheaper than Sonnet.

Context length affects input cost

All Claude models support 200K context. Sonnet 4.5 has 1M beta. If your main bottleneck is input volume rather than reasoning depth, consider whether Sonnet is sufficient before defaulting to Opus.

Pricing summary

All Claude models use per-token pricing. Prompt caching (0.1x cache reads) and batch processing (50% off) are available across all tiers. EvoLink offers discounted access — see each model page for current EvoLink rates.

Opus tier

$5.00 input

$25.00 output

Context: 200K–1M

Flagship tier — use for complex reasoning, coding, and agent tasks where quality is the priority.

Sonnet tier

$3.00 input

$15.00 output

Context: 200K–1M

Balanced tier — recommended default for most production workloads at 40% less than Opus.

Haiku tier

$1.00 input

$5.00 output

Context: 200K

Speed tier — 5x cheaper than Opus, ideal for high-volume and latency-sensitive applications.

Related Claude guides

Use this page to compare models, then visit guides for pricing details, comparisons, and integration.

Claude API Family FAQ

Everything you need to know about the product and billing.

The Claude API family on EvoLink includes all current Anthropic Claude models — Opus 4.7, Opus 4.6, Opus 4.5, Opus 4.1, Sonnet 4.6, Sonnet 4.5, Sonnet 4.0, and Haiku 4.5. All are accessible through one EvoLink API key and endpoint at discounted rates. See the pricing guide for current EvoLink pricing.
Start with Sonnet 4.6 if you need the best everyday balance of intelligence, speed, and cost. If you need maximum quality, use Opus 4.7. If cost or speed is the priority, use Haiku 4.5.
Opus is Anthropic's flagship tier for the hardest reasoning and coding tasks ($5/$25 per MTok). Sonnet is the balanced tier for everyday production workloads ($3/$15 per MTok). In the latest Claude family routes, Sonnet is the lower-cost default while Opus is the premium choice when answer quality matters most.
Claude Haiku 4.5 at $1 input / $5 output per million tokens. It's 5x cheaper than Opus and 3x cheaper than Sonnet, while matching Sonnet 4 on coding benchmarks (73.3% SWE-bench Verified).
Yes. EvoLink provides a single API key and OpenAI-compatible endpoint for all Claude models. You can switch between Opus, Sonnet, and Haiku by changing the model parameter — no separate accounts or keys needed.