Claude vs ChatGPT vs Gemini vs Cursor vs Copilot vs Perplexity
A plain side-by-side look at the major AI assistants and coding tools: what models each runs, what they cost, where they lead, and where they fall short. Covers Claude, ChatGPT, Gemini, Cursor, GitHub Copilot, and Perplexity. The space moves fast; check vendor docs for the latest pricing.
| FEATURE | Claude | ChatGPT | Gemini | Cursor | GitHub Copilot | Perplexity |
|---|---|---|---|---|---|---|
| Claude Opus 4.8 (flagship), Sonnet 4.6 (recommended default), Haiku 4.5 (fastest). Fable 5 and Mythos 5 launched June 2026 but customer access is currently suspended. | GPT-5.5 (flagship), GPT-5.5 Instant (default), GPT-5.4, GPT-5.4 mini/nano, GPT-5.3-Codex. | Gemini 3.5 Flash (default), 3.5 Pro (rolling out, Deep Think), 3.1 Pro, 2.5 Flash-Lite. 2.0 Flash shut down June 2026. | Multi-model: Claude Opus 4.5/4.6, Gemini 3, GPT-5 family, with Auto-select. Composer 2.0 agent. | Multi-model: GPT-5.2 / GPT-5.3-Codex, Claude Sonnet 4.6 / Opus 4.6, Gemini 3 Pro, Grok Code Fast 1. | Routes across models: in-house Sonar, plus GPT-5.4, Claude Opus 4.6, Gemini 3.1 Pro on paid tiers. | |
| 1M tokens on Opus 4.8 and Sonnet 4.6 at flat rates; Haiku 4.5 at 200K. | GPT-5.5 ~1.05M tokens, 128K max output; higher pricing above 272K tokens. | 1M on 3.5 Flash and 3.1 Pro; 3.5 Pro expected 2M. Input ~doubles above ~200K on Pro. | 200K per request with self-summarization for long sessions; model sets hard limits. | Varies by model (up to ~1M on largest); Copilot traces cross-file deps rather than one fixed window. | Inherits the selected model's window; no fixed Perplexity limit. | |
| API per M tokens: Opus 4.8 $5 in / $25 out, Sonnet 4.6 $3 / $15, Haiku 4.5 $1 / $5. Batch 50% cheaper; prompt caching up to 90% off cached input. Consumer: Free, Pro, Max. | API per M: GPT-5.5 $5 / $30, GPT-5.4 $2.50 / $15, Nano $0.20 / $1.25. Cached input 10%; Batch 50% off. Consumer: Free (ad-supported US), Plus $20, Pro $100-$200. | API per M: 3.5 Flash $1.50 / $9, 3.1 Pro $2 / $12, 2.5 Flash-Lite $0.10 / $0.40. Batch 50% off. Consumer free to $249.99/mo; AI Pro $19.99. | Hobby (free), Pro $20 ($20 model credits), Pro+ $60, Ultra $200, Teams $40/user. Credit-based usage; Auto mode unlimited. Bugbot review +$40/user on Teams. | Free; Pro $10 (agent mode), Pro+ $39, Business $19/user, Enterprise $39/user. Monthly Pro/Pro+ move to usage-based AI Credits from June 1 2026. | Free (Sonar, limited Deep Research and Pro Searches); Pro $20 (unlimited Pro Search, ~20 Deep Research/day, frontier models, $5/mo Sonar API credits); Max; Enterprise. Comet AI browser free since March 2026. | |
| Yes, built-in web search | Yes, built-in browsing | Yes, native Google Search grounding | Yes, @Web and @Docs | unverified | Yes, cited multi-source web search is the core product | |
| Yes | Yes | Yes (text, image, audio, video) | Yes (screenshots; agents can drive a browser) | unverified | Yes | |
| No native image generation | Yes, native | Yes, Nano Banana line | No | No | unverified | |
| Voice mode in mobile apps | Advanced Voice on web, iOS, Android | Yes, Gemini Live real-time voice | unverified | unverified | Yes, in the free Comet browser | |
| Strong. Opus 4.8 is positioned as the best coding model; Claude Code is the dedicated agentic CLI; Sonnet 4.6 ~79.6% SWE-bench Verified. | Strong. GPT-5.3-Codex and the Codex agent handle autonomous coding. | Strong. 3.5 Flash leads prior flagship on coding/agentic benchmarks; Antigravity 2.0 agent platform. | Core strength. Agent Mode edits/tests/runs across files; up to 8 parallel agents on isolated branches, Cloud and Background Agents, browser E2E. | Strong. In-IDE Agent Mode plus a cloud coding agent that works from a GitHub Issue, runs tests, opens a draft PR. Copilot CLI with plan/autopilot/fleet modes. | Not a coding tool; Comet does agentic browsing, not software dev | |
| Yes, Anthropic created MCP; deepest support | Yes | Yes; also proposed WebMCP | Yes | Yes, GA in VS Code, JetBrains, Eclipse, Xcode; replaced proprietary extensions | unverified | |
| MCP, Claude Code, tool use, computer use, Agent SDK; on Bedrock, Vertex AI, Microsoft Foundry | GPT Store / custom GPTs, Assistants and Agents APIs, function calling, Codex, operator/computer use. Legacy fine-tuning winding down to new users. | Antigravity 2.0 (CLI, SDK, Managed Agents), Gemini API, AI Studio, Vertex AI, function calling, deep Workspace and Android | MCP, @-mention context, rules, Cloud/Background Agents, Automations, Bugbot | MCP, copilot-instructions.md, Next Edit Suggestions, agentic code review, Copilot CLI, deep GitHub/Actions | Sonar API, Comet AI browser with agentic search and automations, file uploads, Labs | |
| Agentic coding, long-context analysis, production assistants and agents | General-purpose assistant, multimodal creation incl. image gen, voice, coding via Codex | Low-cost multimodal creation, Search-grounded answers, agent-first dev, Google ecosystem | AI-native IDE for multi-file agentic coding, large refactors, parallel cloud agents | Teams on GitHub wanting in-IDE completions plus cloud agents tied to Issues, PRs, Actions | Fast cited research and current-info lookup, plus agentic browsing via Comet |
Claude
- Claude Opus 4.8 (flagship), Sonnet 4.6 (recommended default), Haiku 4.5 (fastest). Fable 5 and Mythos 5 launched June 2026 but customer access is currently suspended.
- 1M tokens on Opus 4.8 and Sonnet 4.6 at flat rates; Haiku 4.5 at 200K.
- API per M tokens: Opus 4.8 $5 in / $25 out, Sonnet 4.6 $3 / $15, Haiku 4.5 $1 / $5. Batch 50% cheaper; prompt caching up to 90% off cached input. Consumer: Free, Pro, Max.
- Yes, built-in web search
- Yes
- No native image generation
- Voice mode in mobile apps
- Strong. Opus 4.8 is positioned as the best coding model; Claude Code is the dedicated agentic CLI; Sonnet 4.6 ~79.6% SWE-bench Verified.
- Yes, Anthropic created MCP; deepest support
- MCP, Claude Code, tool use, computer use, Agent SDK; on Bedrock, Vertex AI, Microsoft Foundry
- Agentic coding, long-context analysis, production assistants and agents
ChatGPT
- GPT-5.5 (flagship), GPT-5.5 Instant (default), GPT-5.4, GPT-5.4 mini/nano, GPT-5.3-Codex.
- GPT-5.5 ~1.05M tokens, 128K max output; higher pricing above 272K tokens.
- API per M: GPT-5.5 $5 / $30, GPT-5.4 $2.50 / $15, Nano $0.20 / $1.25. Cached input 10%; Batch 50% off. Consumer: Free (ad-supported US), Plus $20, Pro $100-$200.
- Yes, built-in browsing
- Yes
- Yes, native
- Advanced Voice on web, iOS, Android
- Strong. GPT-5.3-Codex and the Codex agent handle autonomous coding.
- Yes
- GPT Store / custom GPTs, Assistants and Agents APIs, function calling, Codex, operator/computer use. Legacy fine-tuning winding down to new users.
- General-purpose assistant, multimodal creation incl. image gen, voice, coding via Codex
Gemini
- Gemini 3.5 Flash (default), 3.5 Pro (rolling out, Deep Think), 3.1 Pro, 2.5 Flash-Lite. 2.0 Flash shut down June 2026.
- 1M on 3.5 Flash and 3.1 Pro; 3.5 Pro expected 2M. Input ~doubles above ~200K on Pro.
- API per M: 3.5 Flash $1.50 / $9, 3.1 Pro $2 / $12, 2.5 Flash-Lite $0.10 / $0.40. Batch 50% off. Consumer free to $249.99/mo; AI Pro $19.99.
- Yes, native Google Search grounding
- Yes (text, image, audio, video)
- Yes, Nano Banana line
- Yes, Gemini Live real-time voice
- Strong. 3.5 Flash leads prior flagship on coding/agentic benchmarks; Antigravity 2.0 agent platform.
- Yes; also proposed WebMCP
- Antigravity 2.0 (CLI, SDK, Managed Agents), Gemini API, AI Studio, Vertex AI, function calling, deep Workspace and Android
- Low-cost multimodal creation, Search-grounded answers, agent-first dev, Google ecosystem
Cursor
- Multi-model: Claude Opus 4.5/4.6, Gemini 3, GPT-5 family, with Auto-select. Composer 2.0 agent.
- 200K per request with self-summarization for long sessions; model sets hard limits.
- Hobby (free), Pro $20 ($20 model credits), Pro+ $60, Ultra $200, Teams $40/user. Credit-based usage; Auto mode unlimited. Bugbot review +$40/user on Teams.
- Yes, @Web and @Docs
- Yes (screenshots; agents can drive a browser)
- No
- unverified
- Core strength. Agent Mode edits/tests/runs across files; up to 8 parallel agents on isolated branches, Cloud and Background Agents, browser E2E.
- Yes
- MCP, @-mention context, rules, Cloud/Background Agents, Automations, Bugbot
- AI-native IDE for multi-file agentic coding, large refactors, parallel cloud agents
GitHub Copilot
- Multi-model: GPT-5.2 / GPT-5.3-Codex, Claude Sonnet 4.6 / Opus 4.6, Gemini 3 Pro, Grok Code Fast 1.
- Varies by model (up to ~1M on largest); Copilot traces cross-file deps rather than one fixed window.
- Free; Pro $10 (agent mode), Pro+ $39, Business $19/user, Enterprise $39/user. Monthly Pro/Pro+ move to usage-based AI Credits from June 1 2026.
- unverified
- unverified
- No
- unverified
- Strong. In-IDE Agent Mode plus a cloud coding agent that works from a GitHub Issue, runs tests, opens a draft PR. Copilot CLI with plan/autopilot/fleet modes.
- Yes, GA in VS Code, JetBrains, Eclipse, Xcode; replaced proprietary extensions
- MCP, copilot-instructions.md, Next Edit Suggestions, agentic code review, Copilot CLI, deep GitHub/Actions
- Teams on GitHub wanting in-IDE completions plus cloud agents tied to Issues, PRs, Actions
Perplexity
- Routes across models: in-house Sonar, plus GPT-5.4, Claude Opus 4.6, Gemini 3.1 Pro on paid tiers.
- Inherits the selected model's window; no fixed Perplexity limit.
- Free (Sonar, limited Deep Research and Pro Searches); Pro $20 (unlimited Pro Search, ~20 Deep Research/day, frontier models, $5/mo Sonar API credits); Max; Enterprise. Comet AI browser free since March 2026.
- Yes, cited multi-source web search is the core product
- Yes
- unverified
- Yes, in the free Comet browser
- Not a coding tool; Comet does agentic browsing, not software dev
- unverified
- Sonar API, Comet AI browser with agentic search and automations, file uploads, Labs
- Fast cited research and current-info lookup, plus agentic browsing via Comet
Tool details
Tap a tool to expand strengths, weaknesses, best-fit use cases, and sources.
Related reading
Keep up as the landscape shifts.
One weekly email covering Claude, Claude Code, and the broader agentic AI ecosystem. No hype, no filler.
Subscribe free