Auto Research and Email Research are the same pipeline until the last mile. You branch and iterate in chat, then flip the send-to-inbox switch when you are ready for a formal memo.
What brief are you brave enough to finish before the sun comes up?
Leemer-only model handoffs
No gradients, just the handoff map we run on every research sprint.
GPT-OSS-120B
Triages the vibe of every prompt, tags web or file needs, and preps the branch canvas in under a second.
120B activations
Leemer Deep Research 80B
Qwen3 Next 80B Thinking chews through uploads, diagrams, and citations for the deep reasoning lap.
131K context
Leemer E-Research Pro
Groq-powered scouts fetch sources while Qwen3 Next stitches the email-ready draft and structured citations.
Plan → Cite
MiniMax-M2.1, the new Qwen thinking family, and Kimi K2 Thinking now headline our Leemer lineup.
MiniMax-M2.1 keeps our agentic stack lean and lightning-fast without sacrificing composite intelligence. It ranks at the top of the open-source charts for math, coding, and reasoning workloads.
10B activated params · 204K token contextMoonshotAI’s Kimi K2 Thinking now powers the toughest planning and multi-step reasoning. We tune it to feel like a GPT-5.1 Chat rival, especially on structured debates and multi-query threads.
MoE + 256K context · beats GPT-5.1 Chat in some benchmarksThe latest Qwen release keeps our vision-language bench stacked, adding stronger spatial reasoning and diagram-aware dives alongside our other showcase models.
Multi-modal · newly added to the Leemer tierThere is no separate product. Auto Research is where you live, Email Research is what happens when you flip the “Send as Email” switch inside the exact same chat after the findings are ready.
Keep everything inside the same canvas: branch chats, drop screenshots, upload decks, and swap GPT-OSS, Claude, Grok, or Qwen without losing the thread.
When the branch is locked, toggle “Mail it” in the same session and we package the exact content into a structured email with citations, attachments, and teammate routing.
Pulled straight from /research for clarity:
2023 Launch
9M tokens sparked
9.0M
tokens
2023 Close
40M tokens by winter
40.0M
tokens
2025 YTD
1B+ tokens already
1.0B
tokens
Nine million experimental tokens in early 2023 gave us the data we needed. Forty million tokens by the end of 2023 hardened the pipeline. Surpassing a billion tokens in 2025 is proof that Auto, Email, Writer, and Agents now run on real customer throughput.
Kicks off with GPT-OSS-120B, pivots to Claude or Grok for debates, and finishes with Leemer email-grade drafts without leaving the workspace.
Leemer Deep Research 80B and Kimi K2 Thinking team up for multistep plans, diagrams, and document uploads with 256K+ context.
Keyboard combos, live HTML sandboxes, and drag-in files keep builders shipping prototypes before breakfast.
Bring Groq, Zapier, and custom keys. Auto deploys tokens where it counts while Usage Insights keeps the runway clear.
Shadcn inspired theming, installable desktop, and welcome flows that make every new teammate production ready.
Pick a teammate thread, attach sources, and let the Leemer mailer send curated updates with citations and attachments.
Partner models stay inside the selector below. Here’s the Leemer-native lineup with the Lobe icon treatment you asked for.
Routes GPT-OSS + Claude + Grok
One command fans out to the best brain, keeps Firecrawl feeds alive, and snapshots branches before you swap models mid-thread.
Qwen3 Next 80B Thinking
Multi-hop reasoning for document packs, investor memos, and patent dives. Loves 256K uploads and diagram-heavy briefs.
Groq compound search
Fast reconnaissance engine that visits pages, extracts tables, and preps structured summaries for Auto or Email Research.
Plan → Qwen3 Next 80B
Planning stage, question generator, parallel Groq runs, then Qwen3 Next synthesis. The exact pipeline lifted from the Research surface.
Zapier MCP on Groq
Automation agent that can open tickets, update Notion, and mail the recap. Bring your MCP server URL and watch it coordinate 8k+ apps.
256K dense context
When you need to ingest wikis or legal binders, Heavy keeps the thread stable for hours with steady tone control.
Cognitive telemetry
Live glyphs stream straight from the same counters that power the token timeline, so the shader sphere is a literal look at Auto Research staying in motion.
Orbit speed
0.75c
Shader speed locked to research refresh cadence.
Signal fidelity
99.2%
Letters and numerals mirror the active branch stack.
Tune the exact model stack that fuels Auto Research, Email Research, the writer, and leaderboard challenges. We kept the designer-friendly chrome, now with Lobe icons on every row.
Compare capabilities, context windows, and sweet spots at a glance. Search anything—from providers to features—and we'll surface the model that fits the job.
Quick search
Showing all 48 models
Image generation limits
Free: 10 images/day, 300/month on `black-forest-labs/flux.2-pro`. Paid: unlimited images plus `black-forest-labs/flux.2-flex`, `openai/gpt-5-image-mini`, and `google/gemini-2.5-flash-image`.
Orchestrate automations across 8,000+ apps via Zapier MCP. K2 plans, executes, and audits every step so Gmail, Slack, Google Drive, Notion, Asana, and more stay perfectly in sync.
Paste your Zapier MCP Server URL in Settings → API Keys (or the welcome screen) and K2 discovers available tools, enforces permissions, and remembers preferred playbooks.
Chain Gmail, Slack, Drive, Notion, and hundreds more in a single mission with typed variables and branching logic.
Securely execute Zapier MCP actions with scoped permissions, live status tracking, and automatic retries when steps fail.
Generate summaries, approvals, and audit trails so teammates can step in mid-run without losing context.
Granular logging, environment pinning, and rate protections keep high-impact automations safe and reviewable.
Scope access to just the Zapier tools required for each workflow so reviews stay fast and every action log remains crisp.
Automatically selects the best model for your prompt from a curated set of high-quality, fast models. Powered by OpenRouter's Auto Router, it analyzes your prompt and routes to the optimal specialist—from blazing-fast general models to specialized coding and vision models.
Perfect for everyday use when you want smart routing without the overhead of manual model selection. Just say what you need and Auto (Fast) handles the rest.
Auto scans the vibe of your prompt—coding bug, research sprint, or quick explainer—and snaps to the right specialist in under a second.
Drop in screenshots or photos and Auto pivots to our vision-language lineup so diagrams, charts, and UI mocks get the attention they deserve.
Need raw speed, chain-of-thought reasoning, or long-form analysis? Auto balances the trade-offs automatically—no dropdown spelunking required.
Set Auto as your daily driver and forget the menu. It front-loads great picks so you can stay in flow while still swapping models on demand.
Set Auto as your default model in the chat sidebar. It reads every prompt, routes the moment your cursor leaves the box, and you can still override to a specific model any time.
Stream screenshots, diagrams, or UI mocks and get fast structured breakdowns.
Drop in PDFs and research decks—our lineup ingests long-form context in one go.
Deterministic reasoning models keep agents reliable, verifiable, and auditable.
Our hand-picked lineup for production work: blazing-fast inference, deep context windows, and multi-capability support out of the box.
Qwen
Qwen
Qwen
Qwen
Qwen
Mistral
Mistral
Mistral
OpenAI
OpenAI
OpenAI
OpenAI
OpenAI
OpenAI
OpenAI
Black Forest Labs
Black Forest Labs
MiniMax
ByteDance
ByteDance
Z.AI
Z.AI
Moonshot
Moonshot
xAI
xAI
xAI
Anthropic
Anthropic
DeepSeek
Stepfun
Amazon
Amazon
Xiaomi
Microsoft
NVIDIA
NVIDIA
OpenRouter
OpenRouter
OpenRouter
Long-context multimodal models with native Google search grounding and file understanding.
Low-latency models riding Groq's tensor streaming engine—perfect when every millisecond counts.
Groq
Groq
Groq
Crew, v4.5 is a love letter to every person who DM'd us screenshots, bug repros, and wild feature threads. You asked for clearer boundaries between Auto and Email Research so we opened the pipeline, you wanted our own models front and center so we shipped the Leemer-only lineup, and you kept pushing for better visuals so the new brain sphere is literally powered by shader math.
We are working ridiculous hours to make sure GPT-OSS-120B, Leemer Deep Research 80B, Groq scouts, and Qwen3 Next behave like a single organism. Next up is the Leemer API preview so you can hit one endpoint and get mixture-of-model routing in a single request. It is the same orchestration we run internally, now packaged for your own stacks.
We also moved everyone onto the Autumn pricing track so bills stay predictable while we keep sponsoring open source research, docs, and demo repos that inspire us. When you see a new OSS project carrying the Leemer badge, it is because your subscriptions made it happen.
Need anything, want to test the API preview, or have an idea we missed? Email dev@leemerchat.com and the core team (including me) will answer.
Stay relentless, stay curious, and ship the message no one expects.
Repath Khan
Founder and CEO