Memory
Tell Solan something once and it carries into every chat. You can see, edit, download or delete every note.
Memory settingsEverything you need. Nothing you didn’t ask for.
Open chatReasoning effort
Choose how long Solan thinks before it answers, from an instant reply on Low to the most careful one on Max.
How each level performsWeb search
When a question needs fresh information, Solan searches the web, reads the pages and answers with what it found.
Photos and files
Send photos, PDFs, spreadsheets, notes and code. Solan looks at the images and reads the files alongside your message.
Runs code
Solan writes Python or JavaScript, runs it in a private sandbox and hands back the output, or a file to download.
Searches your past chats
Ask about something you discussed weeks ago and Solan finds the conversation for you.
It can also read your Notion pages and GitHub repositories once you connect them in Settings.
Measured, then published.
All Fresco 1.3 results82.3%HumanEval+ coding
Within a point of Claude Sonnet 3.5 on the harder EvalPlus set.
75.7%MBPP+ coding
Ahead of Claude Sonnet 3.5's 74.3% on the same benchmark.
87/100BFCL V4 tool use
Choosing and calling tools correctly on a function-calling test.
150/150Safety re-test passed
Every adversarial scenario, every run, published in full.
HumanEval+, next to the field
Fresco 1.3 was trained by a small independent lab. On the harder “plus” version of HumanEval it sits alongside models from much larger labs.
Reported scores from the EvalPlus leaderboard, accessed September 2026. Test settings differ between providers, so this is a reported-score comparison rather than one controlled run.
No invisible ink.
Read the commitmentGenerated text can be shaped so it carries a hidden signature, and a conversation can be sold as ad space. Neither shows on the page. Hold the lamp over both sheets.Develop both sheets below.
Sheet AText from elsewhere
A sourdough˙ starter wants a warm⁚ spot and a little patience. Feed it once˙ a day with equal parts flour and water,ˑ and by the fifth day it should double⁚ within six hours.
If it smells˙ like nail polish, it is hungry, not dead.ˑ Feed it, wait, and the sharp⁚ smell fades by the next morning.˙
Sheet BText from Sennoric
A sourdough starter wants a warm spot and a little patience. Feed it once a day with equal parts flour and water, and by the fifth day it should double within six hours.
If it smells like nail polish, it is hungry, not dead. Feed it, wait, and the sharp smell fades by the next morning.
Sheet A carries a statistical watermark and an ad slot. Sheet B is what a Sennoric model returns: the words, and only the words. Why we refuse both
The Solan family
api.sennoric.comprobing…
Default modelChecking…
Solan Fresco 1.4
The default Solan model. It replaced Fresco 1.3 on October 4, 2026, and refuses far fewer harmless questions: on the XSTest benchmark, over-refusal fell from 21.2% to 3.2%. It thinks as long as you ask it to, looks at photos, searches the web, runs code and remembers what you tell it.
- 82.3%HumanEval+
- 87/100BFCL V4 tool use
- 150/150Safety re-test
- 5Reasoning levels
Retired Solan Fresco 1.2.5 Solan Glyph 1.1 Decommissioned September 28, 2026. Why
Build with Sennoric
API reference# a free Sennoric account and API key are required$ curl https://api.sennoric.com/v1/chat/completions \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $SENNORIC_API_KEY" \ -d '{"model": "fresco-1.4", "messages": [{"role": "user", "content": "hello!"}]}'
# pip install openaifrom openai import OpenAI client = OpenAI(base_url="https://api.sennoric.com/v1", api_key="SENNORIC_API_KEY")reply = client.chat.completions.create( model="fresco-1.4", messages=[{"role": "user", "content": "hello!"}],)print(reply.choices[0].message.content)
// npm install openaiimport OpenAI from "openai" const client = new OpenAI({ baseURL: "https://api.sennoric.com/v1", apiKey: process.env.SENNORIC_API_KEY })const reply = await client.chat.completions.create({ model: "fresco-1.4", messages: [{ role: "user", content: "hello!" }],})console.log(reply.choices[0].message.content)
The Sennoric API implements the OpenAI chat-completions format. An existing client needs a new base URL and model name; the rest of the code is unchanged. Streaming behaves as it does with OpenAI.
Browser
sennoric.com/chat
Signed-in conversations sync across devices.
API
POST /v1/chat/completions
Works with the official OpenAI SDKs in any language.
Sennoric CLI
npm i -g @axion-labs-ai/quark-cli
An open-source terminal agent that edits files and runs commands.
From the ledger
All announcements- September 28, 2026Release
Fresco 1.3 Is Now Live
Now the default model everywhere, with doubled weekly limits through October 12 and more context.
Read announcement - September 18, 2026Update
Fresco 1.3 Benchmarks: Matches Claude Sonnet 3.5 on HumanEval+, Beats It on MBPP+
Fresco 1.3's coding and tool-use results on the harder EvalPlus test sets.
Read announcement - September 18, 2026Safety update
Fresco 1.3 Retrained for Safety — 150 of 150 on the Complete Re-Test
Retrained to close the gaps found in August, then re-tested in full: every response passed.
Read announcement
Ask Solan anything.
A free Sennoric account is all it takes. Your words stay your words: nothing hidden in the text, nothing sold in the conversation.