Back to blog
Frontier refresh

Grok 4.3 is in. Old Grok routes are out. Qwen 3.6 just made the lineup cleaner.

We replaced every active Grok route in LeemerChat with x-ai/grok-4.3, promoted qwen/qwen3.6-max-preview into the premium Qwen slot, added qwen/qwen3.6-35b-a3b as the efficient open-weight option, and retired older Qwen 3.5 baggage that no longer deserved UI space.

Apr 20
Kimi K2.6 shipped

Moonshot pushed a 262K-context multimodal agent model built around swarm-style parallel work and open weights.

Apr 27
Qwen 3.6 dropped

Alibaba split the refresh cleanly: one open-weight efficiency route and one proprietary frontier route.

Apr 30
Grok 4.3 arrived

xAI shipped the current flagship with a 1M-token window, text-plus-image input, and cheaper listed pricing than the Grok routes we were carrying.

xAI flagship

x-ai/grok-4.3

1M context$1.25/M in • $2.50/M out

Grok 4.3 is now the only Grok route we surface in LeemerChat. That matters because it collapses the old Grok split between fast, code-fast, and 4.20 beta into one current recommendation with better long-context headroom and cleaner product messaging.

  • Text + image input with reasoning-first behavior
  • Better fit for deep research and large repo sessions than the older Grok variants
  • Lower listed OpenRouter price than Grok 4.20's launch route
Qwen premium

qwen/qwen3.6-max-preview

262K context$1.04/M in • $6.24/M out

Qwen3.6 Max Preview replaces the old Qwen Max / 397B-era premium story with one clear frontier route. It is the expensive, heavy-hitter Qwen pick for agentic coding, structured output, and long-context reasoning.

  • Proprietary sparse MoE with roughly 1T total parameters
  • Built for coding agents, function calling, and multi-turn reasoning traces
  • Cleaner premium Qwen slot than keeping multiple aging 'max' labels alive
Qwen open weight

qwen/qwen3.6-35b-a3b

262K native / 1M via YaRN$0.15/M in • $1.00/M out

Qwen3.6 35B A3B is the pragmatic addition: an open-weight multimodal MoE with 35B total parameters and only 3B active per token. It is far cheaper than Max Preview while still keeping vision, video, structured output, and integrated thinking.

  • Open weight under Apache 2.0
  • Text, image, and video input support
  • The right Qwen option when cost and speed matter more than absolute flagship positioning

Grok 4.3 vs Kimi K2.6

This is not a neat single-number fight. Kimi K2.6 still matters because it is the open-weight, swarm-first coding model with a strong long-horizon builder story. Grok 4.3 matters because xAI finally made the pricing and context math cleaner enough that one current flagship can replace the messy Grok split in a production catalog.

CategoryGrok 4.3Kimi K2.6
Release dateApr 30, 2026Apr 20, 2026
Context window1,000,000 tokens262,142 tokens
PositioningxAI reasoning flagshipOpen-weight multimodal swarm model
Primary strengthLong-context research and agent loopsLong-horizon coding and parallel sub-agent work
OpenRouter pricing$1.25/M in • $2.50/M out$0.74/M in • $3.49/M out

If you want open weights, agent swarms, and the Moonshot coding philosophy, Kimi K2.6 stays compelling. If you want a bigger context window, one current xAI route, and a cleaner product default, Grok 4.3 is the better catalog decision.

Why we replaced the older models

One Grok is easier to trust than three almost-overlapping Groks

The old xAI surface had become catalog debt. Users should not have to decide between a fast route, a code-fast route, and a beta flagship when one newer model now covers the serious use cases better.

Qwen needed a cleaner ladder

The old Qwen lineup mixed dated 3.5 labels with several overlapping 'max' and mid-tier routes. Qwen 3.6 Max Preview plus Qwen 3.6 35B A3B gives us a simple premium-versus-efficient split that actually makes sense.

The frontier is moving too fast for stale labels

April 2026 alone reset the Chinese-model and xAI catalogs. If we keep old names around too long, the app starts teaching users the wrong mental model of the market.

The short version

Grok is now a one-model story in LeemerChat. Qwen now has one clear premium route and one clear efficient route. Kimi K2.6 stays because it solves a different problem: open-weight, swarm-scale coding. That is the frontier in May 2026. Cleaner catalogs are part of product quality too.

Explore models

Related Posts

April 26, 2026

From PowerCode to Critique: The Full Story Behind LeemerChat v8.0.0 and What Happens Next

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 24, 2026

The Trillion-Parameter Open-Weight Wave Hits LeemerChat: DeepSeek V4, MiMo 2.5, Kimi K2.6, and a Limited Ling Promo

We 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 moreApril 4, 2026

We Just Moved a Chunk of the Frontier Into the Free Tier

Three 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 moreMay 6, 2026

Introducing LeemerStudio: Image and Video Generation Built Into LeemerChat

LeemerStudio is a new creative workspace inside LeemerChat for generating images, animating references, rendering video, tracking live status, and keeping every output in private history.

Read more
Explore more:All PostsReleasesModelsBenchmarksEngineeringInsightsAll FeaturesAbout UsTermsPrivacy