Blog

Claude vs ChatGPT vs Gemini: which to pick for coding, writing, research and long documents, plus what the APIs cost

Ramashankar S · · Updated · 15 min read

Claude vs ChatGPT vs Gemini, in short: pick Claude for coding and for working through long documents, ChatGPT for the widest set of tools (Codex, image creation, voice, agents) and the lowest API cost per task, and Gemini for research, writing and anything that lives in Google's apps. All three paid plans cost about $20 a month: Claude Pro $20, ChatGPT Plus $20, Google AI Pro $19.99.

The gaps are smaller than the marketing suggests, so the tie-breakers are practical: which apps you already use, how much text you paste in, and whether you code. Below is one table with the plans, strengths and API prices for all three, a pick for each job, and what it costs to use all three models through their APIs. For a deeper two-way look, see Claude vs ChatGPT and Gemini vs ChatGPT. Every figure comes from the makers' own pages and public leaderboards on 10 November 2026.

Claude vs ChatGPT vs Gemini in one table

ClaudeChatGPTGemini
MakerAnthropicOpenAIGoogle
Free plan modelsSonnet and HaikuA small Luna modelFlash-Lite only
Main paid planPro, $20 ($17 a month billed yearly)Plus, $20Google AI Pro, $19.99
Low-cost planNoneGo, $8 (may show ads)AI Plus, $4.99
Top plansMax $100, Ultra $200Pro $100, $200, $500AI Ultra $99.99, $199.99
Top model on the $20 planClaude Opus 5.5GPT-6 (Sol), up to High thinkingGemini 4 Argon
Context on the $20 planUp to 1M tokens256K reasoning, 54K Instant1M tokens
Coding tool includedClaude CodeCodexAntigravity (entry limits)
Image and video creationNot on the plan pageImagesImages and video
ExtrasCowork, Research, file creationDeep research, projects, custom GPTs, scheduled tasksDeep Research, Gmail and Docs, 5 TB storage
Best atCoding, long documentsTools, API value per taskWriting, research, Google apps
Flagship API priceOpus 5.5 $4 / $20GPT-6.1 Sol $2 / $10Gemini 4 Argon $4 / $20 from 1 Dec ($2 / $10 before)
Low-cost API modelHaiku 5.5 $0.10 / $0.50GPT-6 Luna $0.10 / $0.503.5 Flash-Lite $0.30 / $2.50

API prices are per million tokens, input / output. Sources: Claude pricing, ChatGPT plans and the ChatGPT pricing page, Google AI plans, OpenAI API pricing, Gemini API pricing.

Claude vs ChatGPT vs Gemini compared on price, top model, context window, coding tool and API prices

Three details in that table change the answer for some people:

  • The free plans are not equal. Gemini's free plan gives only the Flash-Lite model with a 32K-token context window, while Claude's free plan includes Sonnet. ChatGPT's free plan has unlimited text chats, but its Instant mode has a 27K-token window.
  • ChatGPT Plus doesn't include the strongest ChatGPT settings. Extra High thinking and GPT-6 Pro (powered by GPT-6 Astra) need a Pro, Business or Enterprise plan, according to OpenAI's model guide.
  • Claude's plan page lists no image or video generation. If you make pictures or clips, ChatGPT or Gemini is the better $20.

The models behind each app in November 2026

Each app runs on its maker's own model family, and all three changed in the last ten weeks:

  • Claude: Opus 5.5 (22 September) is the top model on every paid plan. Sonnet 5.5 (28 September) and Haiku 5.5 (7 October) are the faster tiers. Fable 5.2 launched on 9 November at $8 / $40, down from $10 / $50 for Fable 5.1 (release notes). Pro users pay extra for Fable; Max and Ultra include it up to half the weekly limit.
  • ChatGPT: GPT-6 Sol powers the Instant through Extra High thinking levels, and GPT-6 Astra powers GPT-6 Pro. In the API, GPT-6.1 Sol (29 September) and GPT-6.1 Terra (26 October) are the value models.
  • Gemini: Gemini 4 Argon became available to everyone in October and powers Google AI Pro and Ultra. Google's newer Gemini 4 Carbon tops the Arena text leaderboard, but it is limited to a group of cyber defenders and isn't in the Gemini app or the public API (Google announcement).

On the Artificial Analysis Intelligence Index at each model's high setting, Claude Opus 5.5 scores 52, GPT-6.1 Sol 50 and Gemini 4 Argon 50. Fable 5.2 (62) and GPT-6 Astra (55) sit above them at higher prices. On general ability, the three flagships are close enough that the job decides.

Claude vs Gemini vs ChatGPT for coding

Claude is the pick for coding. Claude Code comes with Claude Pro, and Claude models lead the agentic coding benchmarks.

BenchmarkClaudeChatGPT (OpenAI)Gemini (Google)
Terminal-Bench 4.0 (Vals)Opus 5.5 65.15%GPT-6 Astra 59.60%, GPT-6.1 Sol 55.05%Gemini 4 Argon 57.58%
Cost per task on that runOpus 5.5 $12.04Astra $9.58, Sol $1.72Argon $17.64
WebDev Arena (9 Nov)Fable 5.2 1832, Opus 5.5 1818GPT-6 Astra 1767Carbon 1768 (restricted)
Coding tool on the $20 planClaude CodeCodexAntigravity, entry limits

Sources: Vals Terminal-Bench 4.0 (updated 9 November), Arena WebDev (8 November).

ChatGPT is the value choice: GPT-6.1 Sol solves 55% of the Terminal-Bench tasks for $1.72 each, about a seventh of Opus 5.5's cost, and Codex is included even on the free and Go plans in limited form. Gemini's coding route is Antigravity, Google's agent IDE and CLI, with "entry rate limits" on AI Pro; Gemini 3.8 Flash, the model on Gemini's lower tiers, scores 19.70% on Terminal-Bench 4.0.

The plan limit matters as much as the model. One Hacker News commenter found Codex on ChatGPT Plus used up a five-hour window in minutes during code review (see the evidence section below). OpenAI doubled the Work and Codex allowance on Pro 100 and Pro 500 in late October and says Plus usage is unchanged. The Claude Code pricing guide and the best LLM for coding cover limits and per-task results in detail.

Gemini vs Claude vs ChatGPT for writing

Gemini and Claude are close for writing, and both lead ChatGPT on the one public preference test.

On Arena's creative writing leaderboard (9 November), where people vote blind between two answers, Gemini 4 Argon scores 1525 and Claude Opus 5.5 1518, a gap inside the error margins. The best OpenAI model is GPT-5.6 Sol at 17th place with 1471. On the overall text leaderboard, no OpenAI model makes the top 15.

That is a vote on style, not on facts. In practice:

  • Claude suits long drafts: a 1M-token window on Pro means a whole manuscript, style guide and past drafts fit in one conversation, and file creation turns output into documents.
  • Gemini suits writing that lives in Google's apps: Google AI Pro puts Gemini into Gmail and Docs.
  • ChatGPT suits short-form work with many tools around it: projects, custom GPTs, image creation and memory across chats.

Which to pick for research

Gemini is the pick for research on a small spend, and ChatGPT Pro for the heaviest research. Gemini's Deep Research is on every plan, including free, and Google AI Pro adds a 1M-token window for the sources you upload.

Research featureClaudeChatGPTGemini
Free planNo ResearchLimited deep researchDeep Research included
$20 planResearch includedExpanded deep researchDeep Research, 4x free limits
Top planMax and Ultra: more usagePro: maximum deep researchUltra: 5x or 20x AI Pro limits

Sources: Claude pricing, ChatGPT pricing, Google AI plans.

None of the three publishes a fixed number of research reports per day for every plan, and all three describe limits that reset over five-hour and weekly windows. Treat any report as a draft: check each cited source before you reuse a claim.

Which to pick for long documents

Claude handles the longest documents at the lowest plan price. Every Claude plan lists a context window of up to 1M tokens (it varies by model), which Anthropic puts at about 555,000 words. Gemini matches 1M tokens only from Google AI Pro up, and ChatGPT tops out at 400K on Pro.

Plan levelClaudeChatGPTGemini
FreeUp to 1M (varies by model)27K Instant32K
Low-cost planNo plan54K Instant, 256K reasoning128K (AI Plus)
$20 planUp to 1M54K Instant, 256K reasoning1M
Top plansUp to 1M128K Instant, 400K reasoning1M

Sources: Claude pricing, ChatGPT pricing compare table, Gemini Apps limits.

Context window by plan for Claude, ChatGPT and Gemini, from free to the top plans

ChatGPT notes that the space for your own input is smaller than the total, since instructions, memory and reasoning share the window: about 320 pages of text on Plus with a reasoning model, against about 40 pages in Instant mode. A 300-page contract fits in all three paid plans only if you pick a reasoning mode in ChatGPT.

What it costs to use all three through APIs

To use all three through their APIs, you need three accounts (Anthropic, OpenAI and Google), each billed per million tokens. For one person's chat-style use, the three flagships together cost about $81 a month at full volume each, or $27 if the same work is split three ways. That compares with $59.99 for all three $20 subscriptions.

ModelInput / output per 1MOne person, a monthAn app, 100,000 requests
Claude Opus 5.5$4 / $20$32.40$2,000
Claude Sonnet 5.5$2 / $10$16.20$1,000
GPT-6.1 Sol$2 / $10$16.20$1,000
GPT-6.1 Terra$1 / $5$8.10$500
Gemini 4 Argon$4 / $20 from 1 Dec$32.40$2,000
Gemini 3.8 Flash$0.75 / $3.75 to 31 Dec$6.08$375
Claude Haiku 5.5 or GPT-6 Luna$0.10 / $0.50$0.81$50

The assumptions: one person sends 900 requests a month (30 a day), averaging 4,000 input tokens (the conversation so far plus pasted text) and 1,000 output tokens, reasoning included. The app sends 100,000 requests of 2,000 input and 600 output tokens. No caching and no Batch; Batch halves every figure on all three providers if results can wait up to 24 hours. Redo the math with your own numbers in the token cost calculator.

Monthly API cost for one person on each model, against the $20 subscriptions

Four price details move these numbers:

  • Gemini 4 Argon doubles from $2 / $10 to $4 / $20 on 1 December 2026 (the table uses the new price), and Gemini 3.8 Flash doubles to $1.50 / $7.50 on 1 January 2027, which takes the app column to $750 (Gemini API pricing).
  • OpenAI charges 2x input and 1.5x output on prompts over 272K tokens. Claude Haiku 5.5 moves to $0.50 / $2.50 above 100K.
  • Token counts differ by tokenizer. OpenAI puts 100 tokens at about 75 words, Google at 60 to 80, and Anthropic's figure works out to about 55. The same English text can cost roughly a third more tokens on Claude than on GPT, so add that to the Claude rows before comparing.
  • Cached input costs a tenth or less of the input price on all three (on Opus 5.5, Sonnet 5.5 and GPT-6.1 Sol it is 5%), so a chat app with a long fixed system prompt pays far less than the table shows.

The API makes sense when you build something, or when your use is light: under about 1,100 requests a month at the token counts above, Sonnet 5.5 or GPT-6.1 Sol costs less than a $20 plan. The app plans make sense for heavy daily chat and coding, since they bundle tools such as Claude Code, Codex and Deep Research that you'd otherwise rebuild. The Claude API pricing and Gemini API pricing guides cover caching, Batch and limits per provider.

For research that leans on web search, Perplexity vs Claude compares a search engine that answers with citations against Claude.

What developers report

  • A Hacker News commenter comparing the $20 plans for coding: "Gemini has the most generous limits on the $20 plan." The same comment ranks Claude second and ChatGPT third (Hacker News, 28 Oct 2026, link)
  • Another found Codex on ChatGPT Plus used up a five-hour window on code review in about three minutes, while Claude Pro covered two or three medium-sized features on similar use (Hacker News, 22 Sep 2026, link)
  • A third called Gemini 4 Argon a fantastic model and said picking just one of the three subscriptions would be hard (Hacker News, 29 Oct 2026, link)
  • Arena's blind votes put Gemini 4 Argon first and Claude Opus 5.5 second for creative writing, with no OpenAI model in the top 16 (Arena, 9 Nov 2026, link)
  • Anthropic priced Claude Fable 5.2 at $8 / $40, 20% below Fable 5.1, on its 9 November launch (Anthropic release notes, 9 Nov 2026, link)

Plan limits are the most common complaint across all three, and they change often. Before you commit to a yearly plan, run a week of your real work on the monthly one.

One key for all three APIs

If you build on more than one of these models, Aight gives you claude-opus-5-5 and claude-sonnet-5-5 at Anthropic's list prices, gpt-6.1-sol, gpt-6-astra and gpt-6-luna at OpenAI's, and gemini-4-argon behind one OpenAI-compatible key (https://api.getaight.ai/v1). Switching between the three is a change of model string, and one prepaid balance replaces three accounts. Every recharge adds 10% extra balance, so $1 of list-price usage costs about $0.94 after the 3% processing fee outside India. Each key can have a monthly cap and a list of allowed models.

Where going direct works better:

  • Aight documents chat completions, embeddings and the model list, so there's no Batch discount. For overnight jobs, each provider's own Batch API at half price costs less.
  • Claude Fable 5.2 isn't in the catalog yet.
  • Gemini 3.8 Flash costs less on Google's own API until 31 December, and Google's free tier costs nothing for prototypes.
  • For the apps themselves (Claude, ChatGPT, Gemini, Claude Code, Codex), the subscriptions above are the right route.

FAQ

ChatGPT vs Claude vs Gemini: which is better?

None wins every job. Claude leads coding benchmarks and offers up to 1M tokens of context on every plan; ChatGPT has the broadest tool set and the lowest API cost per task with GPT-6.1 Sol; Gemini ties Claude for writing on Arena's blind votes and includes Deep Research for free. Pick by your main use, then test a month.

Is it worth paying for all three?

For most people, no. All three cost $59.99 a month together, and the flagships score within a few points of each other. Pay for the one that matches your main job, use the free plans of the other two for second opinions, and use the APIs for anything you automate.

Which AI is best for coding: Claude, ChatGPT or Gemini?

Claude. Opus 5.5 scores 65.15% on Terminal-Bench 4.0 against 59.60% for GPT-6 Astra and 57.58% for Gemini 4 Argon, and Claude Code is included with Claude Pro. GPT-6.1 Sol costs about a seventh as much per task at 55.05%, so ChatGPT and Codex are the value pick.

Which has the biggest context window?

Claude and Gemini both reach 1M tokens. Claude lists up to 1M on every plan including free (it varies by model), while Gemini gives 32K free, 128K on AI Plus and 1M from Google AI Pro. ChatGPT's largest window is 400K, on Pro with a reasoning model.

Which is cheapest to use through the API?

At the low end, Claude Haiku 5.5 and GPT-6 Luna both cost $0.10 / $0.50 per million tokens, and Gemini's Flash models have a free tier. Among the flagships, GPT-6.1 Sol at $2 / $10 costs half as much as Claude Opus 5.5 at $4 / $20, and as Gemini 4 Argon once it moves to $4 / $20 on 1 December.

Does Gemini have a free plan with its best model?

No. Gemini's free plan uses the Flash-Lite model only, and Gemini 4 Argon starts with Google AI Pro at $19.99 a month. Claude's free plan includes Sonnet, and ChatGPT's free plan includes a small GPT model with unlimited text chats.


Sources

Checked on 10 November 2026. Updated 17 November 2026: Aight's current model list.