Spring update · April 26, 2026 · 13 min read
If you have been away for a week or two, this is the catch-up. v8 is the release where the partner model row caught up to early 2026, the live search path stopped fracturing across adapters, and the product line finally said out loud that PR-native ship work belongs on Critique.sh, not in a side panel inside LeemerChat.
TL;DR
LeemerChat v8 adds DeepSeek V4 Flash and Pro, MiMo V2.5 and MiMo V2.5 Pro, and Kimi K2.6; standardizes live web search on Firecrawl; and moves review-first coding work to Critique.sh while LeemerChat keeps frontier chat, research, and GitHub-aware help. The shell and settings are cleaner, and the free tier still includes credible frontier models.
Executive summary
v8 is two coordinated moves. First, the product line: LeemerChat stays a frontier workspace for chat, research, and multimodal work, while Critique.sh is the home for pull-request review and builder-native ship loops that used to live under PowerCode. Second, the stack: partner models, search infrastructure, and the app shell all moved forward in the same two-week window so the experience feels like one system, not a patch series.
The app did not step back from ambition. Settings are easier to scan. The editorial shell matches from landing to blog to pricing. The journal is a reliable place to read what merged. The product reads as a point of view, not a feature backlog with a model modal bolted on.
| Theme | What changed |
|---|---|
| Partner models | DeepSeek V4 Pro and Flash, MiMo V2.5 Pro and V2.5, Kimi K2.6, and related metadata and spend controls, alongside the existing frontier stack (for example GPT-5.4, GLM-5, Claude, Gemini, MiniMax, Qwen). |
| Web search | Firecrawl is the standard live search path; the same layer feeds extraction and research-heavy flows, not a forked stack. |
| Code and review | In-app PowerCode is retired; Critique.sh is the product for PR-native review. LeemerChat still uses GitHub context for normal chat and coding help. |
| App shell | Modular settings, stronger mobile behavior, a simplified tips surface, and an editorial pass across major marketing and product pages. |
Shipping history
Eight merges between mid-April and the v8 line are the concrete reason the version number moved. Summaries are drawn from the same work that powers the public release notes.
DeepSeek V4 Pro and Flash, MiMo V2.5 Pro and V2.5, Kimi K2.6, and the InclusionAI Ling promo ship across model selectors, metadata, spend controls, and blog surfaces so the roster and pricing story stay in sync.
Landing, Foundry, and research pages were rebuilt with a calmer grid, better assets, and a single through-line: what the product is for, not a feature laundry list.
Settings now uses a modular shell: quick actions up front, account and model management in one rhythm, and less visual noise between panes.
Touch targets, scroll containment, and viewport behavior in the settings dialog were tightened so the same flow works on phone-sized screens without jank.
The old tips surface was pared back so it can track the current release; settings picked up the companion cleanup to match.
About, blog, marketing pages, header, footer, pricing, and the updates feed share one voice and spacing system again.
A syntax error in the post registry is gone, which matters now that the journal is a first-class place to read what shipped.
LeemerLabs shows up as a real company and product story in-app, on its own page, in the sitemap, and in product update hooks so new visitors can see the lab connection.
Model refresh
The most visible change is the roster. DeepSeek V4 Flash and Pro, Xiaomi MiMo V2.5 and MiMo V2.5 Pro, and Kimi K2.6 sit next to GPT-5.4, GLM-5, GLM-5V Turbo, Claude, Gemini, MiniMax, Qwen, and the rest of the frontier bench. The partner row is not just newer names: several entries are large-context, agent-weight models, and some of them still appear on the free path, which changes what “default chat” can mean.
The bet is that users will pick models the way they pick compute: by task fit, not by a single flagship slot. v8 makes that bet easier to act on in the product.
DeepSeek V4 Pro
The full DeepSeek V4 profile for long-horizon reasoning, serious coding, and multi-step agent runs when latency is a fair trade for depth.
DeepSeek V4 Flash
The speed-oriented sibling: better for tight feedback loops, high-volume chat, and agent tool loops that need answers back quickly.
MiMo V2.5 Pro
Xiaomi’s 1M-context flagship, positioned as a peer to other frontier options rather than a budget filler.
MiMo V2.5
A lower-cost omnimodal line that still reads as a deliberate pick, not a degraded default.
Kimi K2.6
Moonshot’s long-context, multimodal step up for heavy synthesis, document work, and orchestration across tools.
Yes. The gap between free and paid is more about how hard you can lean on the stack than whether you are allowed to use serious models at all.
Because the product needed one honest answer to “where does live web search run?” The codebase now routes modern search and research tooling through Firecrawl as the default substrate.
You can trace it in the dedicated web-search path, the main tool registry, and the deeper research stack where search, scrape, extract, browser, and crawl operations are all first-class tools.
In practice, free and paid search-enabled chats lean on the same backbone. Plans still differ where it matters: model orchestration, throughput, and how deep a workflow is allowed to go. Standardizing the substrate is what makes that distinction legible to users instead of feeling like a lottery of which adapter you hit.
Because review-native, repo-shaped work outgrew a secondary surface inside a general chat app. v8 is the line where LeemerChat stops straining to host that workload inline.
PowerCode showed that teams wanted serious engineering work next to their models. It also showed that pull-request review, builder runs, and durable code artifacts do not want to live forever as a side panel. Critique.sh is the product that follows from that lesson instead of fighting it.
LeemerChat is not leaving code behind. The main app still ships with GitHub context, codebase-aware answers, and top-tier coding models. The change is only where review-first ship work is supposed to start. The long-form narrative lives on the PowerCode to Critique post.
The best version of the platform is not one window that awkwardly does everything. It is a sharp frontier workspace for chat, research, and multimodal work, with adjacent products that own their own jobs. The recent PR history reads like the team internalized that.
v8 is worth a detailed post because the badge is not decorative: the product line is clearer, the model stack is stronger, search has a single spine, and the split between LeemerChat and Critique is finally explicit in the UI story, not only in release engineering.
Short answers for people skimming, aligned with the same facts as the v8.0.0 release notes and partner copy on /tips.
LeemerChat v8.0.0 retires the in-app PowerCode agent (4.9+). What it was, why Critique.sh fits review and ship work better, and what stays in main chat. Partner terms on /tips.
Read moreApril 9, 2026Z.AI's GLM-5.1 is the first model built for long-horizon autonomous coding — running independently for 8+ hours, planning, executing, and self-improving without human input. It beats or matches GPT-5.4 and Claude Opus 4.6 on several benchmarks. Here's why we made it free, how it compares, and why Pro still matters.
Read moreApril 4, 2026Three frontier-grade models go free — Xiaomi MiMo-V2-Pro (1M context), Z.AI GLM-5V-Turbo (native multimodal agent), and Google Gemma 4 31B IT (89.2% AIME 2026). Premium gets sharper with GPT-5.4, GLM-5, and MiniMax M2.7. Plus: why we cleaned up the lineup and what frontier actually means now.
Read moreApril 24, 2026We refreshed our partner lineup around Chinese-led trillion-scale open models—DeepSeek V4 Pro and Flash, Xiaomi MiMo 2.5 Pro and 2.5, Moonshot Kimi K2.6, plus a time-bound free route for InclusionAI Ling 2.6 1T. Here is why we bundled them, how they differ, and how to benchmark them on real work instead of leaderboard screenshots alone.
Read moreThree parallel AI research agents for citation-ready reports and follow-ups. Free deep research alternative.
Try itPowerCode in LeemerChat is retired. GitHub review and builder workflows continue on Critique.sh (LeemerChat / Leemer Group).
Try it