3.7 Flash or Sonnet 5
Intro rate vs a locked $2/$10
claude-sonnet-5 is callable on QCode; QCode does not currently offer the Gemini series, so the gemini-3.7-flash figures are Google's official data, for comparison only. 3.7 official intro is $0.75 / $3.75 through 31 Dec 2026, then $1.50 / $7.50. Sonnet 5 confirmed $2 / $10 as standard on 10 Aug 2026 (the planned Sep $3 / $15 was cancelled). This page does not invent independent head-to-heads.
Updated 2026-10-08
One key across Claude, GPT and Chinese models, billed per token.
- Billed per token — live rates on /models
- Pay by card (Visa / Mastercard / AMEX), Apple Pay, Google Pay or crypto
- Self-service: top up and activate instantly
What actually compares
vs $2 / $10
3.7 is an intro rate through 31 Dec 2026; from 1 Jan 2027 it is $1.50 / $7.50. Sonnet 5’s $2 / $10 is now standard. Google’s own launch table also lists Sonnet 5 at $2 / $10 — label that vendor.
vs 1M / 128K
Official 3.7 Flash: 1,048,576 in, 65,536 out. Sonnet 5: 1M context, 128K out. Same order of window; Sonnet allows a larger completion.
vs Sonnet thinking
Those are 3.7’s thinking levels. Officially, minimal errors. Do not paste these names onto Sonnet 5.
The side callable on QCode
QCode does not currently offer the Gemini series; claude-sonnet-5 is on /models, and claude-sonnet-5-5, gpt-6.1-sol or glm-5.3 are alternatives. Live list wins.
What we are comparing
A Flash workhorse against Claude’s mid-tier default. Six weeks apart: Sonnet 5 on 30 Jun 2026, 3.7 Flash on 13 Aug 2026. Price, window, output cap and thinking controls line up with official docs. Independent head-to-head reruns are not invented here.
Google’s own comparison table
The DeepMind model card / launch table lists Claude Sonnet 5 at $2 in / $10 out and prints vendor figures (DeepSWE v1.1: 3.7 at 65.3% vs Sonnet 5 at 53.8%; FrontierCode 1.1 Main: 43.6% vs 42.7%). Those rows are Google’s table, not a rerun on this site, not an independent rank. Sonnet 5 at $2 / $10 matches what Anthropic confirmed as standard on 10 Aug 2026.
Timeline
Claude Sonnet 5 ships. 1M / 128K.
Anthropic confirms $2 / $10 as standard. Planned 1 Sep $3 / $15 is cancelled. See /claude-sonnet-5-pricing.
Gemini 3.7 Flash GA. Intro $0.75 / $3.75 through 31 Dec 2026. Google’s table includes Sonnet 5 at $2 / $10.
Confirmed vs do not invent
Confirmed
Both ids are in their official docs. claude-sonnet-5 is callable on QCode; QCode does not currently offer the Gemini series. 3.7 intro $0.75 / $3.75 through 31 Dec 2026, then $1.50 / $7.50. Sonnet 5 standard $2 / $10. Windows and output caps as above. Google’s launch table lists Sonnet 5 at $2 / $10.
Do not invent
Do not publish a homemade “independent bake-off.” The scores above are Google’s table and must be labeled vendor. Do not annualize 3.7’s intro rate. Do not still write that Sonnet 5 jumps to $3 / $15 in September.
When to pick which
Pick 3.7 Flash
You want the lower list price, Flash-class throughput, official low / medium / high, and you can live with the 1 Jan 2027 double and a 65,536 output cap.
Pick Sonnet 5
You want the Claude stack, 128K out, a standard $2 / $10 that is no longer a promo, or a workflow only validated on Sonnet.
One key, swap the id
On QCode set model to claude-sonnet-5 (or claude-sonnet-5-5, gpt-6.1-sol or glm-5.3); QCode does not currently offer the Gemini series. When calling 3.7 through Google's own API, use low / medium / high — not minimal. For Sonnet 5 use Anthropic’s own thinking interface; do not reuse Google’s level names.
On QCode
claude-sonnet-5 is on /models and callable on QCode. QCode does not currently offer the Gemini series; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3). Billing is per token, with each model's price on /models.
3.7 Flash vs Sonnet 5 FAQ
Which is cheaper?
List: 3.7 intro $0.75 / $3.75, Sonnet 5 standard $2 / $10. From 1 Jan 2027, 3.7 becomes $1.50 / $7.50. Real bills still depend on completion length and thinking tokens.
How different is the window?
3.7: 1,048,576 in, 65,536 out. Sonnet 5: 1M context, 128K out. Input is the same order; Sonnet allows a larger completion.
Who wins at coding?
We do not invent an independent rank. Google’s own table prints DeepSWE v1.1 65.3% vs 53.8% and FrontierCode 1.1 43.6% vs 42.7%. Label those vendor.
Can I call both on QCode?
No. claude-sonnet-5 is callable on QCode, but QCode does not currently offer the Gemini series; please use Claude, GPT or Chinese models (e.g. claude-sonnet-5-5, gpt-6.1-sol, glm-5.3).
How do I set thinking?
3.7: low / medium / high; minimal errors. Sonnet 5 uses Anthropic’s own thinking interface — do not paste these three names onto it.
What should be the default?
Stay on Sonnet 5 if the workflow is already locked to Claude. Try 3.7 if you want the Flash list price and can accept a 65,536 output cap. Regress first; do not trust the vendor table alone.
Sources
Google Keyword / DeepMind model card, 13 Aug 2026 (vendor table listing Sonnet 5 at $2 / $10); Gemini API pricing; Anthropic Sonnet 5 docs and the 10 Aug 2026 standard-price confirmation, see /claude-sonnet-5-pricing.
Want both? Change the model field
claude-sonnet-5 is callable on QCode; the Gemini series is not currently offered, so use GPT or Chinese models instead.
Related
3.7 Flash guide
Specs, intro rate, how to call it.
Claude Sonnet 5
Mid-tier default. 1M / 128K, $2 / $10.
Sonnet 5 vs GPT-5.6 vs Gemini 3
The three-way page. A different table.
Not affiliated with Google or Anthropic. Vendor tables are not independent reruns. Prices follow each vendor’s docs and /models.