Quick verdict. There is no single winner. Claude Sonnet 4.6 leads Truth Score at 96 against Gemini 3 Flash at 80 and Gemini 3.1 Pro at 84, and leads privacy at 94. Flash leads speed at 95 and API price at $0.50/$3. Pro matches Sonnet on task at 89 with a million-token window. Match the job, not the logo.
Gemini versus Claude is a business-job comparison, and Gemini is not one row in the dataset. The workhorse Google model is Gemini 3 Flash, ranked 4 with a weighted score of 86. The frontier Google model is Gemini 3.1 Pro, ranked 6 with a weighted score of 85. The Claude model in this head-to-head is Claude Sonnet 4.6, ranked 2 with a weighted score of 87. Task fit, truth and API cost do not crown the same side. Flash task performance is 78 against Sonnet's 89. Pro task performance is 89, level with Sonnet, on a 1,000,000-token window. Truth Scores are 80 for Flash, 84 for Pro and 96 for Sonnet. API rates are $0.50 and $3 for Flash, $2 and $12 for Pro, and $3 and $15 for Sonnet, all per million input and output tokens. Privacy is 72 for Flash, 74 for Pro and 94 for Sonnet. HIPAA is false only on Flash. None of the three is self-hostable. Jurisdiction for all three is the United States. Use the sections below to match the job, not the logo.
Name the models before you pick a side
This is not the ChatGPT vs Claude page with the vendor swapped. Gemini has two relevant rows. Flash is ranked 4, weighted 86. Pro is ranked 6, weighted 85. Claude Sonnet 4.6 is ranked 2, weighted 87. None is self-hostable. All three are marked United States jurisdiction.
| Lens | Gemini 3 Flash | Gemini 3.1 Pro | Claude Sonnet 4.6 |
|---|---|---|---|
| Weighted score | 86 | 85 | 87 |
| Task | 78 | 89 | 89 |
| Truth Score | 80 | 84 | 96 |
| API input / output per million | $0.50 / $3 | $2 / $12 | $3 / $15 |
| Context window | 1,000,000 | 1,000,000 | 200,000 |
| Speed | 95 | 80 | 82 |
| Privacy | 72 | 74 | 94 |
| HIPAA | false | true | true |
Figures from /data/ai-comparison-2026.json. Last verified 2026-06-26. Scoring version v1.0. Affiliate links are off.
Task fit
Sonnet and Pro both sit at task 89. Flash sits at 78. Coding and writing jobs should start with Sonnet's best-fit list and the coding guide. Support bots and high-volume classification are on Flash's list and on customer service. Pro's list is Google Workspace integration, long documents, multimodal work and data analysis, which is a different job from Flash's queue.
Truth
The gap is large: 96 against 84 and 80. If the business job fails when a fact is invented, Sonnet is the published leader of this trio. Truth method is on Truth Score. A truth lead is not a citation checker. Unsourced footnotes still fail the citation rule.
API cost
Flash is the cheap side of this pair at $0.50 and $3 per million tokens, cost efficiency 89. Pro is $2 and $12, cost efficiency 74. Sonnet is $3 and $15, cost efficiency 76. Run the volume through the calculator before you treat the sticker as the decision. Profiles: Gemini 3 Flash and Claude Sonnet 4.6.
Privacy breaks the tie for regulated jobs
Privacy is 94 on Sonnet, 74 on Pro and 72 on Flash. HIPAA is false only on Flash. For medical-record text, Flash is out on that flag alone. Provider pages, not affiliate links: gemini.google.com and anthropic.com/claude.
Match the constraint you actually have. The match engine takes task, budget and privacy as separate questions.
FAQ
Which Gemini model is this comparison?
Both rows in the dataset. Gemini 3 Flash is the speed and cost workhorse. Gemini 3.1 Pro is the frontier model with task 89.
Who leads on Truth Score?
Claude Sonnet 4.6 at 96, ahead of Gemini 3.1 Pro at 84 and Gemini 3 Flash at 80.
Who is cheaper at the published API rates?
Gemini 3 Flash at $0.50 input and $3 output per million tokens. Sonnet is $3 and $15. Pro is $2 and $12.