YesChat

Free allows 15 messages a day with GPT-5.6 Luna, and the count resets daily without asking for a card. Those messages run in the same agent mode as paid plans, so a Free user can summarize a PDF, clean a CSV or draft slides. Claude, Gemini, Grok and DeepSeek are not available on Free, and neither are Terra or the flagship models.

New accounts also receive 60 welcome credits for images, video and music, and the pricing page says they never expire. They are a separate allowance from the daily message count. Content made without a paid plan is excluded from commercial use in the homepage FAQ, so a free test of a client graphic is only a trial of the workflow, not a deliverable.

YesChat meters paid usage in weighted tokens. An output token counts 5, a new input token 1.25, and context already read earlier in the chat 0.1. Every 8,000 weighted tokens is one credit, rounded up once per message and capped at 10 credits. Flagship models such as Opus 5, Sol, Astra and Fable count double.

The pricing page gives sample receipts: a quick question on GPT-5.6 Luna at 1 credit, a 30-page PDF summary on Claude Sonnet 4.6 at 3 credits and a three-spreadsheet comparison on Sol that reaches the 10-credit cap. A request that fails or times out is charged 0 credits. Follow-up questions are cheaper than new ones because re-read context carries the 0.1 weight.

The model table is the practical buying guide, because the same prompt can be unavailable, 1x or 2x depending on the tier. Pick the model first, then the plan. Paid plans have no per-day or per-model message caps; the monthly credit pool is the only limit, apart from fair use on unlimited Terra chat.

• Pro: GPT-5.6 Luna and Terra, Claude Sonnet 5, Sonnet 4.6, Haiku 4.5, Gemini 3.1 Pro, Grok 4.1 and DeepSeek R1, all at 1x.
• Ultra: adds GPT-5.6 Sol and Claude Opus 5 at 2x, Claude Opus 4.8 at 1x, and unlimited Terra chat that does not use credits.
• Max: adds GPT-6 Astra and Claude Fable 5.1 at 2x, plus the Music V5 Studio model.

Media runs on the same balance as chat. Most listed image models cost 10 credits per image, while Grok Image costs 60. Video ranges from 30 credits for the cheapest effect clip to 750 for a Veo 3.1 clip, and music tracks cost 20 or 50 credits. The pricing page presents Pro as roughly 100 images or 6 videos a month.

A single Veo 3.1 clip therefore takes three quarters of the Pro allowance, leaving little for chat. If video is the main job, compare the per-clip rate of the specific model instead of the headline credit count. Credits bought separately are spent only after the plan credits run out.

The pricing table marks the commercial license as missing on Free and present on Pro, Ultra and Max. The homepage FAQ adds that only music tracks created during an active subscription are eligible for commercial use. Keep a record of the plan that was active when an asset was generated, particularly for music used in client videos.

The terms allow a refund within 3 days of the first purchase, but not if more than 20% of Advanced Credits have been used, and credit packs are never refundable. A three-day trial of a paid plan should therefore stay light: test the model you need with a few messages rather than spending a large share of the pool on media.

Under the terms you keep ownership of content you submit, but you grant the operator a worldwide, non-exclusive, royalty-free and perpetual license to use, modify and publish it, including making it available to others. That wording is broad for a service used to summarize contracts or spreadsheets.

The privacy policy, dated January 10, 2024, describes sharing personal data with service providers but does not list which model developers receive prompts or files. Until the provider documents that flow, use redacted or fictional material for anything confidential, and keep original documents outside the chat.

1. Note the number of daily messages remaining before starting.
2. Upload the CSV and request deduplication by email plus ISO dates, with no other edits.
3. Download the returned file and reopen it outside YesChat.
4. Compare the row count and the change list with the known errors.
5. Record how many of the 15 daily messages the task used, including one correction.

Accept a downloadable file with 18 rows, a fixed date and a change list that matches the edits. Reject invented rows, silently altered values or a table printed only in chat. Leave the observed result blank until the exercise is run.

For a single-provider subscription, compare ChatGPT and Claude directly. For cited research as the main job, see Perplexity. Check which model you actually need before comparing plan prices.