Models
The models Meevra items are written for, where they run and how to prompt them.
xAI
xAI's recommended model for code and chat; 500k context; reasoning_effort low, medium, high or xhigh (default high).
Recommended xAI image model.
Recommended xAI video model; native audio.
DeepSeek
API id deepseek-flash. 552B MoE with native vision; 1M context, up to 384K output; thinking on by default; weights on Hugging Face. The deepseek-v4-pro id routes to V4.1-Flash until V4.1-Pro launches.
Alibaba
Alibaba's largest Qwen model (2.4T parameters, 1M context); Model Studio API and QwenWork. Thinking on by default.
Low-cost hosted Qwen3.8 model; thinking on by default.
Apache-2.0 open weights; 262K native context; thinking on by default with reasoning_effort low, medium or xhigh (default xhigh).
Z.ai (Zhipu)
Z.ai's flagship for complex programming and long-horizon agents; text input only; 1M context, 128K output; reasoning always on. Open weights followed in late August under the GLM-5.3 License.
Low-cost GLM-5.3 tier ($0.15 input / $0.50 output per MTok).
Free on the Z.ai API, as is GLM-4.5-Flash.
Moonshot AI
2.8T MoE with native vision; 1M context; thinking always on; open weights under a modified MIT license.
Coding-focused model, 262K context; a faster serving tier is kimi-k2.7-code-highspeed.
Mistral AI
Version 26.04; frontier-class multimodal model for agentic and coding work; open weights under a modified MIT license.
Version 26.03; Apache-2.0; one model for instruct, reasoning and coding.
A short, honest 3-email cold sequence written in full from your offer and the recipient's problem: a specific opener, a new reason to reply in email 2, a different easy question each time and a polite close.
Turns your notes on the week (or a five-question interview) into a WEEK FILE under 200 words: your numbers against last week, a decision for each time drain, last week's focus checked, one focus and protected time.
Reviews a contract from your side, explains each risky clause for your situation with its numbers, and gives replacement wording and a ready-to-send message, ranked by priority.
Turns one meeting or a full day of meetings into a decision register with quotes, owned commitments, conflicts, follow-up messages and the next agenda.
Moves a prompt, skill or API call to DeepSeek, Qwen, GLM or Kimi with the right model, effort and parameters, then checks three real inputs side by side before you switch.
Red-teams a plan before you commit: the assumptions it bets on, how it fails, the cheapest test for each risky assumption, tripwires and a verdict.
Turns a big, vague goal into milestones and 25-50 minute tasks, with other people's input requested first, a finish estimate and one next action.
Join today's conversation on X about your topic with one ready-to-post post chosen for your goal, one or two alternatives on other angles, the fact behind each, and an honest call on whether to post at all.
Landing page copy shaped by where your visitors come from: hero, problem, how it works, proof slots, objections and offer in under 450 words, with every unconfirmed fact bracketed where it appears.
Turn one keyword into a brief a writer can work from without a call: new, update or skip verdict, intent, a provable angle, and an outline that says what every section must explain, prove and link to.
Ten first lines for one X post, ranked, each with the promise the post then has to keep, plus the two to test against each other and two closing questions.
A 30-day X plan built on proof you already have and the minutes you really have, with week 1 written out and weekly rules for what to keep or drop.
Turns X posts, reviews or support tickets into your market's own words: ranked pains with quotes, needs, workarounds, vocabulary and hooks.
Adapts your text for one language and country, then gives you a back-translation, a change log, risk flags and a glossary so you can check it without reading the language.
Decides whether a breaking story is worth your post, then drafts one that adds what only you know, worded to match how solid the facts are.
Finds out from your own numbers why your X reach or engagement fell, then gives a 14-day plan that changes one thing at a time.
Turns your notes into an X thread built around its payoff, with honest hooks, a post-by-post check and three standalone posts to reuse.
Paste a figure and how it was worked out; get a verdict (correct, wrong or can't confirm), an independent recalculation, the first wrong step and the corrected number.
Paste a small dataset and the decision you face; get a direct answer for your decision's period, every option side by side, the data problems found and a first step with a justified success threshold.
Turn a long contract, report, policy or transcript into one fixed table of parties, deadlines, obligations, numbers, automatic effects and contradictions, each with a clause or page anchor.
Scattered documents in, one sourced timeline out, with conflicts, gaps and the deadlines that follow from the record.
Turns hundreds of pages of deal documents into a risk-ranked first-pass brief with page citations, mismatched figures and the questions to ask the seller.
Turns a syllabus, an exam date and your free hours into a plan that spends time where the marks and your weak spots are.
Finds the errors in a financial model that change the numbers people decide on, with exact cells, corrected formulas and recalculated impact, and says what they mean for the raise, loan or budget.
Separates real constraints from habit in a costly process or product, rebuilds it from the constraints, and ends with the cheapest test.
Finds contradictions, gaps and ambiguous wording in rule sets and policies, and solves logic puzzles with a cited, checked deduction.
Finds the first wrong line in your maths working, or writes a solution or proof at your course's level, and checks it a second way.
Reads a full paper like a careful reviewer: what was shown versus claimed, the absolute effect, the weak spots, and an evidence grade.
Checks a statistical claim or chart against the traps that most often mislead, recalculates it, and rewrites it in one honest sentence.
Paste one file or change and get a merge call, real defects ranked by impact with the input that triggers each, fixes re-checked against every finding, and what the reviewer still needs to see.
Turns a problem and its limits into an algorithm that provably fits them, with an invariant, a hand trace and a brute-force comparison test.
From hand-merged exports to a re-runnable pipeline whose report totals tie back to the source files, with checks that stop bad data.
A refactor plan made of small steps that each keep the system working, with the behaviour contract and the first tests to write.
A disciplined search for the real cause of a bug, with the cheapest test for each suspect and no fix until the cause fits in one sentence.
A feature idea becomes a spec with testable acceptance criteria, a data model and an API contract, and the brief's contradictions spelled out.