Pinned Tweet
KAT-Coder-Pro V2.5 is now live — the next leap in agentic coding.
- Strong Long-Horizon Task Handling
- Excellent General Agentic Capabilities
- Multi-Harness-Driven RL Scaling
API KEY: streamlake.ai/product/kat-co…
Tech report: arxiv.org/abs/2607.05471
KAT-Coder-V2.5-Dev is now open-weight.
Built on Qwen3.6-35B-A3B and post-trained with the proven KAT-V2.5 recipe—from data construction to the SFT/RL pipeline and optimization strategy.
35B total, 3B active. Apache 2.0
↓ Explore the model
Hugging Face: huggingface.co/Kwaipilot/KAT…
The result: strong Agentic Coding performance among models at a similar scale.
• 69.4 SWE-bench Verified
• 63.0 SWE-bench Multilingual
• 45.96 SWE-bench Pro
• 41.02 Terminal-Bench 2.1
The post-training pipeline combines:
• 127K-example supervised fine-tuning
• Reinforcement learning with execution feedback
• Reliable sandboxes and verifiers
• Hierarchical rewards for meaningful progress
• Targeted penalties for failed, excessive, or repetitive tool use
RL training also significantly reduced abnormal agent behavior:
• Abnormal tool labels: 9.34% → 0.28%
• Single-turn continuous repetition: 0.34% → 0%
Heads up devs 🚀 KAT-Coder-Pro v2.5 is free to use on @kilocode for this week only.
blog.kilo.ai/i/207303310/new…
2
20
1,069
🚀 KwaiKAT lands on ZenMux — now 40% OFF!
KAT-Coder-Pro V2.5 is a flagship Agentic Coding model that goes beyond writing code to handling real engineering tasks. 🔥
🛠️ Tackles repo-level issues
🤖 Built for complex agentic workflows
🎯 Strong across coding, tool use, and UI tasks
Now joining the ZenMux Token Economics lineup at 0.6x pricing 👉
zenmux.ai/kuaishou/kat-coder…
#ZenMux #TokenEconomics #KwaiKAT
General agentic capability, strengthened.
V2.5 is built for longer tool-use chains and real workflows such as data analysis, cross-system integration, batch document processing, and report generation.
Not just calling tools — completing workflows.
Long-horizon engineering, upgraded.
V2.5 is trained to handle real repository tasks: understand natural-language issues, navigate unfamiliar codebases, make cross-file changes, follow project conventions, run tests, and recover from failures.
From completing code to running projects.
V2.5 delivers strong results across coding and agentic benchmarks:
- SWE-Bench Pro: 65.2
- KAT Code Bench: 53.1
- PinchBench: 94.2
- KAT Claw Bench: 85.5
Built for both long-horizon repo engineering and multi-step agentic workflows.
Take a guess 👇
Which is KAT-Coder-Pro V2, which is Kimi K2.6?
Drop your answer in the comments
2
9
1,410
KAT + Hermes Practical Case Showcase
Step 1: Install the html-ppt skill
Step 2: Input your specific prompt
Step 3: Get a perfectly matched output in no time
1
9
1,109
Prompt:Design an Apple-style product launch deck for our new AI assistant. 20 slides, cyberpunk-neon theme, pure black background. Cover slide: just the product name with gradient text and a starfield animation. Feature slides should use neural-net FX. Heavy weight contrast — bold titles, thin body. Smooth magic-move transitions throughout.
536
Shoutout to @ArtificialAnlys — We’ll keep pushing forward
KwaiKAT has released KAT-Coder-Pro V2, a non-reasoning model that scores 44 on the Artificial Analysis Intelligence Index, an 8 point improvement from KAT-Coder-Pro V1
@KwaiAICoder has updated their flagship proprietary coding model with the release of KAT-Coder-Pro V2. KAT-Coder-Pro V2 achieves 44 on the Artificial Analysis Intelligence Index, matching Claude Sonnet 4.6 (non-reasoning) and trailing only Claude Opus 4.6 (non-reasoning, 46) among non-reasoning models. At ~9M output tokens, it is also more token efficient than Claude Opus 4.6 (~11M), Claude Sonnet 4.6 (~14M), and reasoning models with similar intelligence such as DeepSeek V3.2 (reasoning, ~61M) and Qwen3.5 397B A17B (reasoning, ~86M).
KAT-Coder-Pro V2 is a non-reasoning model, unlike all of the current frontier language models which ‘think’ before answering. Typically, reasoning variants score higher on the Intelligence Index than their non-reasoning counterparts, but consume more output tokens and are less suited to latency-sensitive workloads.
Key Highlights:
➤ 🧠 Higher overall intelligence, but regression in long context reasoning and knowledge recall: KAT-Coder-Pro V2 scores 44 on the Artificial Analysis Intelligence Index, an 8 point improvement from KAT-Coder-Pro V1 and matching Claude Sonnet 4.6 (non-reasoning, max effort). It performs well on tool use (90% on Tau2-Telecom), but regresses compared to KAT-Coder-Pro V1 on long-context reasoning and knowledge, falling 8 p.p. on AA-LCR (66%) and 17 p.p. on HLE (16%).
➤ 🤖 Agentic capability improvements: KAT-Coder-Pro V2 shows major improvements on our agentic evaluations. On Terminal-Bench Hard, it scores 49%, up 40 p.p. from KAT-Coder-Pro V1, making it the highest-scoring non-reasoning model, matching Claude Opus 4.6 (non-reasoning, 49%) and ahead of Claude Sonnet 4.6 (non-reasoning, 46%). KAT-Coder-Pro V2 also shows improvement in GDPval-AA, scoring 1123 (+304 Elo from V1), but still sits behind models such as DeepSeek V3.2 (1198) and Qwen3.5 397B A17B (1202).
➤ ⚙️ High token efficiency: KAT-Coder-Pro V2 is a non-reasoning model and uses fewer tokens than peers with similar intelligence. It uses 8.7M output tokens to run the Artificial Analysis Intelligence Index, below Claude Opus 4.6 (non-reasoning, ~11M) and Claude Sonnet 4.6 (non-reasoning, ~14M), though this is ~2x higher than its predecessor, KAT-Coder-Pro V1 (~4.5M). It also uses significantly fewer tokens than similarly intelligent reasoning models such as DeepSeek V3.2 (reasoning, ~61M) and Qwen3.5 397B A17B (reasoning, ~86M).
➤ $ Improved cost efficiency: KAT-Coder-Pro V2 costs $73 to run the Artificial Analysis Intelligence Index, down from $76 for V1, as it uses fewer input tokens by requiring fewer turns in agentic evaluations. This makes it one of the most cost-efficient models at its intelligence level, costing less than Qwen3.5 397B A17B (reasoning, $418) and Claude Sonnet 4.6 (non-reasoning, $1397). KAT-Coder-Pro V2 is currently priced at $0.30/$1.20 per 1M input/output tokens on StreamLake and AtlasCloud API endpoints.
➤ ⚡ Low end-to-end response time: KAT-Coder-Pro V2 runs at ~109 output tokens per second, far ahead of Claude Opus 4.6 (non-reasoning, 39 OTPS) and Claude Sonnet 4.6 (non-reasoning, 43 OTPS). Because it also has a low time to first token without any reasoning delay, it delivers one of the fastest end-to-end response times, which measures the time taken from request sent to final output returned.
Model details:
➤ Availability: KAT-Coder-Pro V2 is available via StreamLake and AtlasCloud API endpoints
➤ Context Window: 256K tokens (equivalent to KAT-Coder-Pro V1)
➤ Multi-modal capabilities: Text input and output only
6
34
2,236
Intelligence vs. Price
KAT-Coder-Pro V2 lands in the most attractive quadrant in AA evaluations, combining strong performance with highly competitive pricing in real-world development workflows.
1
1
583
KAT-Coder-Pro V2 is fully available.
API Key:
streamlake.ai/product/kat-co…
Coding Plan:
streamlake.ai/marketing/codi…
1
533















