APIs & SDKs

Amazon Q Developer

✓

Agentic AWS-aware coding assistant with IDE, CLI, and console coverage

Amazon Q Developer (formerly CodeWhisperer) is AWS’s generative AI assistant for coding, application upgrades, and cloud operations. It provides agentic chat in IDEs, the AWS Console, chat apps, and the CLI (upgradeable to the Kiro CLI) to accelerate coding, run secure code scans and tests, and transform Java/.NET applications with guided upgrade workflows. The perpetual Free tier includes 50 agentic requests and 1,000 lines of code transformations per month; Pro is $19/user/month with higher limits, pooled upgrade quotas, admin controls, and IP indemnity.

• ●Agentic chat and coding in IDEs with inline suggestions and repository-aware context
• ●CLI command completions and agentic workflows (Q Developer CLI users can upgrade to Kiro CLI)
• ●AWS Console assistant for troubleshooting, cost optimization, and architecture guidance

Claude 3.5 Sonnet

✓

Prior-generation Claude Sonnet model with 200K context for coding workflows

Claude 3.5 Sonnet (launched Aug 28, 2025) is Anthropic's prior-generation mid-tier Sonnet model line. It was widely used for coding and vision tasks with a 200K context window and $3/$15 per million input/output token pricing. Anthropic's current Sonnet default is the 4.6 line, so verify active model availability and pricing on official pages before implementation planning. Claude Sonnet 3.5 remains a historical benchmark entry and is superseded in many workflows by newer Sonnet 4.6 releases.

• ●200K token context window with text and image inputs
• ●Strong coding and multi-step reasoning performance
• ●Tool use/function calling and structured outputs

Claude 4.5 Haiku

✓

Anthropic's speed-optimized budget model: 200K context, extended thinking, and $1/$5 per MTok

Claude 4.5 Haiku (released November 2025) is Anthropic's speed-optimized budget model, delivering near-Sonnet 4 performance at one-third the cost. With a 200K token context window, extended thinking, and full tool support (code execution, computer use, bash, text editor, web search), it is the fastest option in the Claude lineup. Haiku is available across Claude.ai (web, iOS, Android), Claude Code, the Claude API, Amazon Bedrock, Google Vertex AI, and Azure AI Foundry at $1/$5 per million input/output tokens. It serves as the ideal choice for high-volume, latency-sensitive applications where cost efficiency matters.

• ●200K token context window with text and image inputs (extended thinking supported)
• ●Near-Sonnet 4 performance with faster latency and $1/$5 MTok pricing
• ●Full Claude tool set: code execution, bash, computer use, text editor, web search

Anthropic Claude API

✓

Build with Claude Opus 4.6, Sonnet 4.6, and Haiku via the Anthropic API with MCP and adaptive thinking

The Anthropic Claude API provides pay-per-use access to the full Claude model lineup: Opus 4.6 (released Feb 5, 2026) for highest-depth reasoning at $5/$25 per MTok, Sonnet 4.6 (Feb 17, 2026) as the balanced default at $3/$15, and Haiku 4.5 for speed-optimized budget workloads at $1/$5. The API supports the Model Context Protocol (MCP) with 75+ enterprise connectors, adaptive thinking for extended reasoning, context compaction for efficient long conversations, tool use, prompt caching (write/read discounts), batch processing (50% savings), and deployment across Anthropic, AWS Bedrock, Google Vertex AI, and Azure AI Foundry.

• ●Claude model lineup: Opus 4.6 ($5/$25), Sonnet 4.6 ($3/$15), Haiku 4.5 ($1/$5)
• ●Model Context Protocol (MCP): open standard with 75+ enterprise connectors
• ●Adaptive thinking: extended reasoning mode for complex problem-solving

Claude Opus 4.6

✓

Anthropic's frontier reasoning model: 80.9% SWE-bench record, 1M token beta context, and adaptive thinking

Claude Opus 4.6, released February 5, 2026, is Anthropic's top-tier model for complex coding and agentic workflows. It holds the SWE-bench Verified record at 80.9% and features adaptive thinking for extended reasoning, a 200K standard context window with 1M token beta, and 128K max output tokens. Priced at $5 per 1M input tokens and $25 per 1M output tokens, it powers Claude Code's Agent Teams for autonomous multi-file development and supports 30+ hour task endurance for long-running sessions.

• ●80.9% SWE-bench Verified: world record for autonomous coding benchmarks
• ●Adaptive thinking: extended reasoning mode for complex problem-solving
• ●200K standard context with 1M token beta window

DeepSeek Coder

✓

Open-source MoE coding model (V2) with 128K context

DeepSeek Coder V2 is an open Mixture-of-Experts coding model (16B Lite and 236B) with a 128K context window and support for 338 programming languages. Weights are published on Hugging Face for self-hosting; code is MIT-licensed and the model weights use the DeepSeek Model License. Hosted inference is available through the DeepSeek API using OpenAI-compatible endpoints.

• ●DeepSeek-Coder-V2 Mixture-of-Experts models (236B total / 21B active, plus 16B Lite / 2.4B active)
• ●128K context window with fill-in-the-middle support
• ●Supports 338 programming languages

Google Gemini API

✓

Access Gemini 3 Flash (~78% SWE-bench at $0.50/$3), 3.1 Pro, and the full Gemini model family

The Gemini API exposes Google's full model lineup including Gemini 3 Flash (~78% SWE-bench at $0.50/$3 per MTok, the best price-performance coding model available), Gemini 3.1 Pro Preview (released Feb 19, 2026, at $2/$12), Gemini 2.5 Pro ($1.25/$10), and Gemini 2.5 Flash ($0.30/$2.50). All models support 1M token context windows, multimodal inputs (text, image, audio, video), and tools including function calling, code execution, Google Search grounding ($14 per 1K queries), URL context, and computer use. Google AI Studio provides free prototyping with Build Mode for rapid app development.

• ●Gemini 3 Flash: ~78% SWE-bench at $0.50/$3 per MTok (best price-performance)
• ●Gemini 3.1 Pro Preview: latest flagship (Feb 19, 2026)
• ●1M token context windows across all models (largest in the industry)

Gemini 3 Pro (with 3.1 Pro Preview & 3 Flash)

✓

Google's Gemini 3-series: 3.1 Pro Preview flagship, 3 Flash at ~78% SWE-bench, and 1M token context

The Gemini 3 series is Google's latest multimodal model lineup. Gemini 3 Flash is the default model across Google products, delivering ~78% SWE-bench Verified and 90.4% GPQA at just $0.50/$3 per MTok, making it the best price-performance coding model available. Gemini 3.1 Pro Preview (released Feb 19, 2026, at $2/$12 per MTok) is the newest flagship. Gemini 3 Pro achieves 76.2% SWE-bench but is discontinuing March 2026. All models support 1M token context windows, multimodal input, and tool-grounded workflows for coding and planning.

• ●Gemini 3 Flash: ~78% SWE-bench at $0.50/$3 per MTok (best price-performance)
• ●Gemini 3.1 Pro Preview: latest flagship released Feb 19, 2026
• ●Gemini 3 Pro: 76.2% SWE-bench (discontinuing March 2026)

OpenAI API

✓

API access to GPT-5.2, Codex models, Responses API, Agents SDK, and the full OpenAI platform

The OpenAI API provides pay-per-use access to the GPT-5 family (GPT-5.2 flagship at $1.75/$14 per MTok, GPT-5 at $1.25/$10, GPT-5 Mini at $0.25/$2, GPT-5 Nano at $0.05/$0.40), Codex-oriented models for agentic coding, reasoning models (o3, o4-mini), and GPT-4.1 with 1M token context. The Responses API (launched March 2025) replaces the older Assistants API with built-in tools including web_search, file_search, and computer_use. The Agents SDK provides a Python framework for building multi-agent systems with handoffs, guardrails, and tracing. Also includes image generation (DALL-E), audio (TTS/STT), embeddings, and the Sora 2 video generation API.

• ●GPT-5.2 flagship (71.8% SWE-bench), GPT-5, GPT-4.1 (1M context), o3/o4-mini reasoning
• ●Responses API with built-in web_search, file_search, and computer_use tools
• ●Agents SDK: Python framework for multi-agent systems with handoffs and guardrails

GPT-4o

✓

OpenAI flagship GPT with text + image inputs and 128K context

GPT-4o (“o” for omni) is OpenAI’s fast, high-intelligence GPT model for general tasks. It accepts text and image inputs and returns text outputs (including structured outputs) via the Chat Completions API. It supports cached prompts, batch processing, fine-tuning, and snapshots, with a 128K context window and up to 16,384 output tokens. Knowledge cutoff: Oct 1, 2023. OpenAI announced GPT-4o retirement timelines in favor of newer GPT-5 family paths; confirm current availability windows in official model docs.

• ●128K token context window with up to 16,384 output tokens
• ●Text and image inputs with text outputs and structured outputs
• ●Fast, high-intelligence flagship GPT optimized for general tasks

GPT-5

✓

OpenAI's first unified reasoning model: 70.1% SWE-bench, 400K context, and $1.25/$10 per MTok

GPT-5 is OpenAI's unified reasoning model launched June 2025, designed for coding and agentic tasks. It achieves 70.1% on SWE-bench Verified with text and image inputs, a 400K context window, and up to 128K output tokens. GPT-5 supports reasoning tokens with configurable effort, runs across Chat Completions, Responses, Realtime, and Assistants APIs, and includes tools for web search, file search, image generation, and code interpreter. While GPT-5.2 is now the default ChatGPT model, GPT-5 remains available for production workloads at $1.25/$10 per MTok.

• ●400K token context window with up to 128K output tokens
• ●70.1% SWE-bench Verified for autonomous coding tasks
• ●Text and image inputs with reasoning token support

Perplexity

✓

AI answer engine with real-time citations, Deep Research, and the Sonar API

Perplexity is an AI-powered answer engine that delivers real-time web search results with source citations on every response. Processing 780M monthly queries for 30M active users, it offers Quick and Pro search modes, Focus modes (Academic, Writing, Math, Video), and Deep Research for multi-step autonomous synthesis across dozens of sources. The Sonar API provides developer access to search-augmented AI at up to 1,200 tokens/second via a Cerebras partnership. Perplexity also ships Comet, a free AI-native Chromium browser, and Perplexity Computer, a full autonomous computer agent launched February 26, 2026.

• ●Real-time web search with source citations on every answer
• ●Deep Research for autonomous multi-step synthesis across dozens of sources
• ●Focus modes: Academic, Writing, Math, and Video

Firebase Studio (formerly Project IDX)

✓

Firebase Studio cloud IDE with Gemini-assisted development

Project IDX has been folded into Firebase Studio, an agentic browser IDE with Gemini in Firebase for coding, debugging, testing, and refactoring. The preview is no-cost with 3 workspaces per user; Google Developer Program members get 10 (Standard) or 30 (Premium). App Prototyping and Code Assist agents, built-in previews/emulators, Git imports, and App Hosting deployments are included; some Firebase services bill separately once linked to Cloud Billing.

• ●Agentic browser workspace with Gemini in Firebase chat and code actions
• ●App Prototyping agent to scaffold apps from text, sketches, or templates
• ●Import from GitHub/GitLab/Bitbucket or local; customize via Nix

Vercel AI SDK

✓

Open-source TypeScript toolkit for streaming LLM apps with providers, tools, and generative UI

Vercel AI SDK (v6 beta) is a free, open-source toolkit for building streaming AI apps across Next.js, React, Vue/Nuxt, SvelteKit, and Node/Edge runtimes. It provides a unified provider API, tool calling, structured outputs (`generateObject`), server actions/RSC hooks, and generative UI helpers like `streamUI`, with adapters for OpenAI, Anthropic, Google, Mistral, Ollama, and more.

• ●Unified provider API with adapters for OpenAI/Anthropic/Google/Mistral/Ollama and more
• ●Streaming text, chat, and tool calls with React/Next server actions and RSC hooks
• ●Structured outputs via `generateObject` with schema validation