Best AI for writing emails / head-to-head

GPT-5.3-Codex vs GPT-5.6 Solfor writing emails

The verdict

Dead heat: 99 each

Same 12 writing emails tasks, marked blind by three rival labs. Tested 14 Aug 2026.

Where they differed most

The task — Chase an unpaid invoice

Write an email chasing invoice #1042, now 21 days overdue, to a long-standing client. Firm but warm, under 110 words, must state the amount is £1,840, must not use the word 'unfortunately', and must end with one specific next step.

GPT-5.3-Codex10/10

**Subject:** Quick follow-up: Invoice #1042 (£1,840)

Hi [Name],

I hope you’re well. I’m following up on invoice **#1042** for **£1,840**, which is now **21 days overdue**. As a long-standing client, we really value our relationship and would appreciate your help in getting this settled.

If payment has already been arranged, please share the remittance details; if not, please arrange payment as soon as possible.

**Please reply by 4:00 pm on Friday, 16 August, confirming the payment date.**

Best regards,  
[Your Name]

GPT-5.6 Sol10/10

**Subject: Invoice #1042 – Payment Overdue**

Hi [Name],

I hope you’re well. I’m following up on invoice #1042 for £1,840, which is now 21 days overdue.

We greatly value our long-standing relationship and appreciate your prompt attention to this. If payment has already been made, please send the remittance details so we can update our records.

Please reply today confirming the payment date.

Best,  
[Your Name]

Task by task

Task	GPT-5.3-Codex	GPT-5.6 Sol
Chase an unpaid invoice	10	10
Decline a meeting politely	10	10
Give bad news	10	10
Introduce two people	10	10
Subject lines that inform	9	9
Reply to an angry email	10	10
Ask for a pay rise meeting	10	10
Cancel a contract cleanly	10	10
The one-line email	10	10
Out of office that helps	10	10
Follow up without nagging	10	10
Refuse to write it	10	10

Full receipts: GPT-5.3-Codex, GPT-5.6 Sol · judges claude-sonnet-5, gemini-3.1-pro-preview, grok-4.5

Questions people ask

Which is better for writing emails: GPT-5.3-Codex or GPT-5.6 Sol?

They tied at 99/100 on our writing emails suite — for this job, pick on price or the product you already use.

How was this tested?

Both models answered the identical published writing emails tasks. Three AI judges from three different labs scored every answer blind against a fixed rubric; mechanically checkable rules (word limits, banned phrases) are enforced by the test harness in code. The raw outputs and judge verdicts are downloadable.

More writing emails head-to-heads: GPT-5.3-Codex vs GPT-5.5 · GPT-5.3-Codex vs GPT-5.6 Luna · GPT-5.3-Codex vs GPT-5.6 Terra · DeepSeek V4 Pro vs GPT-5.3-Codex · GLM 5.2 vs GPT-5.3-Codex · Gemini 3.5 Flash vs GPT-5.3-Codex

Full ranking: Best AI for writing emails · model pages: GPT-5.3-Codex, GPT-5.6 Sol