ChatGPT vs Claude vs Gemini Comparison

At-a-Glance Comparison

Note (13 June 2026): Anthropic has disabled Fable 5 (and Mythos 5) for all users to comply with a US government export-control directive restricting foreign-national access. Opus 4.8 is currently the top accessible Claude model; the table reflects Fable 5's specs for reference. Update (2 July 2026): The US government lifted the export controls on 30 June, and Anthropic restored Fable 5 for users globally from 1 July 2026 (Claude.ai, the Claude Platform, Claude Code and Cowork); Mythos 5 is also being re-enabled. Fable 5 is once again the top-tier Claude model. Update (26 June 2026): OpenAI previewed the GPT-5.6 family (Sol / Terra / Luna) as a government-restricted limited preview to a small set of partners; GPT-5.5 remains OpenAI’s top generally available model, so the table reflects 5.5. Wider GPT-5.6 access is expected in the coming weeks. Update (8 July 2026): The US government lifted the preview restrictions on 8 July 2026, and OpenAI has confirmed GPT-5.6 (Sol / Terra / Luna) is launching publicly this week — from Thursday 9 July 2026 — reaching general availability across the API and ChatGPT over the following weeks. We’ll refresh the table to the Sol / Terra / Luna tiers once they are generally available. Update (10 July 2026): GPT-5.6 (Sol / Terra / Luna) reached general availability on 9 July 2026 across ChatGPT, the API and Codex. Pricing per 1M tokens: Sol (flagship) $5 / $30, Terra (balanced) $2.50 / $15, Luna (fast) $1 / $6. GPT-5.6 Sol is now OpenAI’s top model, and free and Go users default to GPT-5.6 Terra; the header above is updated accordingly. Update (19 July 2026): Anthropic’s promotional free access to Fable 5 for Pro, Max, Team and premium Enterprise seats ended at 11:59pm PT on 19 July 2026, after three extensions. Fable 5 remains available but from 20 July 2026 draws on usage credits at its standard $10 / $50 per million tokens rather than the included weekly allowance; Opus 4.8 and Sonnet 5 remain included on paid subscriptions, and Sonnet 5 is the default model on Free and Pro. Anthropic says it intends to return Fable 5 to subscriptions once it has more compute. Note also that Opus 4.8, Sonnet 5 and Fable 5 all now carry a 1M-token context window at standard pricing — the table is updated accordingly. Correction (21 July 2026): Anthropic confirmed that from 20 July 2026 Fable 5 is not removed from subscriptions outright. It is now a standard inclusion on Max, Team Premium and legacy premium Enterprise seats at up to 50% of weekly usage limits. Only Pro and Team Standard move to usage credits at $10 / $50 per million tokens, and those users receive a one-off $100 credit first. Opus 4.8 and Sonnet 5 remain included across paid plans, with Sonnet 5 the default on Free and Pro. Update (24 July 2026): Anthropic released Claude Opus 5, a near-frontier model that reaches roughly Fable 5–level intelligence at half the price and sets new state-of-the-art results on coding and knowledge-work benchmarks. Standard pricing is $5 / $25 per million tokens (the same as Opus 4.8, and half of Fable 5’s input price), with an optional Fast mode at $10 / $50 running about 2.5× faster. Opus 5 adds a per-request reasoning-effort setting (low / medium / high) so you can dial cost against depth. It succeeds Opus 4.8 as the included mid-tier Claude model on paid plans; the header above now reflects Opus 5. Update (28 July 2026): Google made Gemini 3.6 Flash generally available on 21 July 2026 — a 1M-token workhorse that is faster and cheaper than 3.5 Flash (around $1.50 / $7.50 per million tokens), now the default Flash model; the header above is updated accordingly. Gemini 3.5 Pro remains delayed, so 3.1 Pro is still the top Pro tier. For reference, xAI’s Grok 4.5 (public from 9 July 2026, roughly $2 / $6 per million tokens) also sits in the “Opus-class, faster and cheaper” band alongside these models. Update (30 July 2026): OpenAI cut GPT-5.6 API prices as enterprises push for clearer returns on AI spend. Luna drops about 80% to $0.20 / $1.20 per million tokens (from $1 / $6) and Terra falls roughly 20% to about $2 / $12; Sol’s pricing is unchanged at $5 / $30. OpenAI credited efficiency gains from GPT-5.6 rewriting and optimising its own inference stack. Update (14 August 2026): Google made Gemini 3.7 Flash generally available on 13 August 2026, just three weeks after 3.6 Flash. It targets coding, agentic workflows and document processing and posts clear benchmark gains over 3.6 Flash. Google launched it with an introductory price of $0.75 / $3.75 per million tokens (input / output) through 31 December 2026 — roughly half the prior Flash pricing, and applied retroactively to 3.6 Flash — after which list prices rise to $1.50 / $7.50 from 1 January 2027. Gemini 3.7 Flash is now the default Flash model; the header above is updated accordingly. Gemini 3.5 Pro remains delayed, so 3.1 Pro is still the top Pro tier. Update (16 August 2026): xAI released Grok 4.6 on 12 August 2026, holding pricing flat at $2 / $6 per million tokens (the same as Grok 4.5, with a 500K-token context window). It posts 61 on the Artificial Analysis Intelligence Index — matching GPT-5.6 Sol and sitting about two points behind Claude Opus 5 — and is strongest on knowledge work and reasoning, keeping it in the “Opus-class, faster and cheaper” band noted above. Its pitch is price-to-intelligence rather than outright leadership; Grok 4.6 supersedes the Grok 4.5 reference above. Update (21 August 2026): OpenAI cut GPT-5.6 Sol API pricing to $4 / $20 per million tokens (input / output) — down from $5 / $30, a 20% input and 33% output reduction — as a promotional rate running through at least 21 November 2026. The cut applies to the API and to credits on ChatGPT Work and Codex; Pro, Plus and Business subscription pricing is unchanged. OpenAI framed the move as undercutting Claude Opus 5 amid rising competition from Anthropic and Chinese models. Update (1 September 2026): Anthropic released Claude Fable 5.1 (with the trusted-access Mythos 5.1) on 1 September 2026 — the same model at two safeguard levels. Fable 5.1 is generally available and sets a new bar for coding, knowledge work and long-running tasks, at an estimated 25% lower cost than Fable 5 for typical token-billed workloads; its cybersecurity safeguards now block about 60% fewer false positives. It supersedes Fable 5 as the top-tier Claude model; the header above is updated accordingly. Update (2 September 2026): Google introduced Gemini 3.8 Flash (and 3.8 Flash Cyber) on 2 September 2026 — the third Flash release in six weeks — at the same introductory price as 3.7 Flash ($0.75 / $3.75 per million tokens through 31 December 2026), with clear gains over 3.7 Flash on software-engineering, agentic and multi-step reasoning benchmarks. It is now the default Flash model; the header above is updated accordingly. Gemini 3.5 Pro remains delayed, so 3.1 Pro is still the top Pro tier. Update (3 September 2026): Two housekeeping notes. First, since 6 August 2026 free and Go ChatGPT users default to GPT-5.6 Luna (not Terra); Plus and Pro default to GPT-5.6 Sol, where the old Instant and Thinking entries have been folded into a single model with a reasoning-effort slider — the header above now reflects this. Second, Meta released Muse Spark 1.3 on 2 September 2026 (1M-token context; $1.25 / $4.25 per million tokens), which edges Claude Opus 5 and GPT-5.6 Sol on some agentic-coding benchmarks and joins the “Opus-class, faster and cheaper” band alongside Grok 4.6 and Gemini 3.8 Flash. Meta’s 30B open-weight Muse Glimmer (Apache 2.0) is the pick if you need a capable model running locally. Update (3 September 2026, evening): OpenAI released GPT-6 Astra, its new flagship, on 3 September 2026. API pricing is $10 / $50 per million tokens (cached input $1; prompts over 272K input billed 2×; Batch and Flex at 50%) — the same as Fable 5.1 and 2.5× Sol’s promotional rate — with a 1.05M-token context, 128K max output, a knowledge cutoff of 30 April 2026 and five reasoning-effort levels from low to max. It is rolling out to ChatGPT Plus, Pro, Business and Enterprise, the API (gpt-6-astra) and AWS over the coming days; on launch day only Daybreak trusted-access enterprises had it. GPT-5.6 Sol remains the everyday default on Plus and Pro (Luna on Free), so the header above now reads Astra / Sol. Two caveats: Astra is the first OpenAI model over the Preparedness Framework “Critical” cyber threshold, so the public version refuses exploit discovery and advanced cyber work is gated behind Daybreak; and OpenAI says its own monitors may slow, pause or stop legitimate long-running tasks at launch — and that Astra’s chain of thought is harder to monitor than Sol’s. Its launch benchmarks are self-reported at max effort; treat them as a starting point, not a verdict.

When to Use Each Model

Choose ChatGPT When...

Choose Claude When...

Choose Gemini When...

Task-by-Task Recommendations

Quick Decision Framework