AI Radar
A weekly overview of the AI models available: which one to use for
which job, what it costs, and what is new.
This week saw practical developments in software and security, highlighted by a critical remote code execution vulnerability named http2shell affecting Node and Python, alongside discussions on using LLMs for formal software verification at a larger scale. For your homelab, evaluate your Node and Python deployments to address the http2shell RCE risk, and test out local agents for use cases like Pokémon Showdown or multi-user access control.
issue 2026-W43 · generated 2026-10-24
· 3 new models
The short answer #quick
One pick per job: the best value model we found that is also popular
enough to be worth trusting.
| Job | Model | Price | Why |
|---|
| Cheap & small (task routing) |
Xiaomi: MiMo-V2.6-Flash
xiaomi/mimo-v2.6-flash
|
$0.168/1M
|
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting co…
|
| Medium (real work & coding) |
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)
google/gemini-3.1-flash-lite-image
|
$0.500/1M
|
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is an extremely cost-efficient option at $0.500/1M blended tokens, making it great for high-v…
|
| Expensive (architecture) |
OpenAI: GPT-5.4
openai/gpt-5.4
|
$5.00/1M
|
The openai/gpt-5.4 model delivers a unified frontier architecture suitable for heavy homelab automation and complex coding workflows at a price of $5.…
|
| Coding & SWE agents |
Xiaomi: MiMo-V2.6-Flash
xiaomi/mimo-v2.6-flash
|
$0.168/1M
|
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting co…
|
| Out of the box (native tools) |
OpenAI: GPT-5.6 Luna
openai/gpt-5.6-luna
|
$0.400/1M
|
openai/gpt-5.6-luna is a fast and cost-efficient option at $0.400/1M blended tokens with a strong arena rating of 1457. It is well-suited for high-vol…
|
| Voice & audio |
Google: Lyria 3 Pro Preview
google/lyria-3-pro-preview
|
$0.08 per song
|
google/lyria-3-pro-preview is part of Google's family of music generation models available through the Gemini API and is well-suited for generating fu…
|
| Open weights (self-hostable) |
Xiaomi: MiMo-V2.6-Flash
xiaomi/mimo-v2.6-flash
|
$0.168/1M
|
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting co…
|
| Image & video |
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)
google/gemini-3.1-flash-lite-image
|
$0.500/1M
|
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is an extremely cost-efficient option at $0.500/1M blended tokens, making it great for high-v…
|
Cheap & small (task routing) #router
Cheap enough to call on every step. Routing, classification, extraction, summarising.
#1
$0.168/1M · $0.00196 per task
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting cost of $0.168/1M blended tokens. It is well-suited for processing large codebases and documents in a homelab, though its lower popularity score of 46 means community integration support may be lacking. It is not worth the hardware overhead if your local setup cannot efficiently handle its 309B total parameter mixture-of-experts footprint.
cheapopen weightslong context10 providers
#2
$0.220/1M · $0.00250 per task
Z.ai: GLM 5.3 Flash delivers strong efficiency for coding and long-horizon agent tasks with an arena rating of 1472 and a 1.0M context window. At $0.220/1M blended tokens, it is a capable choice for homelab operators who need a native multimodal model backed by widespread provider availability. However, it may not be worth the cost if your local tasks do not require multimodal capabilities or long-context reasoning.
cheapopen weightslong context38 providers
#3
$0.343/1M · $0.00359 per task
Z.ai: GLM 5.3 is built for complex software engineering and long-horizon agent tasks, boasting an arena rating of 1474 and a popularity score of 58 across a 1.0M context window. At $0.343/1M blended tokens, it is the most expensive option in this lineup, making it worth the money only when you need deep reasoning capabilities for intricate codebases. If your homelab workloads consist of simple scripts or routine automation, the higher cost is unnecessary.
cheapopen weightslong context48 providers
Medium (real work & coding) #workhorse
The models that do the actual work: writing, refactoring, debugging, agentic loops.
#1
$0.500/1M · $0.00550 per task
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is an extremely cost-efficient option at $0.500/1M blended tokens, making it great for high-velocity developer pipelines and rapid-fire visual exploration. However, with an arena rating of 1313, it may not be worth the money if your homelab tasks require top-tier visual fidelity.
mid-priceimage out
#2
$0.522/1M · $0.00609 per task
xiaomi/mimo-v2.6-pro costs $0.522/1M blended tokens, backed by a strong arena rating of 1491 and a popularity score of 38. With a massive 1.1M context window and over 1T parameters, it excels at pushing the capability ceiling for the most demanding workloads. It is not worth the money, though, if you only need a lightweight option for simple, low-stakes text tasks.
mid-priceopen weightslong context5 providers
#3
$0.512/1M · $0.00576 per task
qwen/qwen3.7-plus is available for $0.512/1M blended tokens, achieving an arena rating of 1455 and a popularity score of 41. It is a cost-effective choice that supports text and image inputs within a substantial 1M context window. It is not worth the money if your homelab pipeline relies strictly on smaller context limits or non-multimodal text setups.
mid-pricelong context
Expensive (architecture) #architect
When the answer has to be right and the reasoning has to hold up. Use sparingly.
#1
$5.00/1M · $0.0550 per task
The openai/gpt-5.4 model delivers a unified frontier architecture suitable for heavy homelab automation and complex coding workflows at a price of $5.00/1M blended tokens. It provides a massive 1.1M context window, making it a strong choice when you need to process large repositories, but it may not be worth the money for simpler tasks given its high cost.
premiumopen weightslong context7 providersapply patchweb search
#2
$7.20/1M · $0.0800 per task
The anthropic/claude-opus-5.5 model serves as Anthropic's flagship option for demanding reasoning and multi-step changes in large codebases with an arena rating of 1516. Priced at $7.20/1M blended tokens, it is best reserved for intensive agentic work where top-tier accuracy is required, and is definitely not worth the money for routine or low-stakes queries.
premiumlong context11 providersadvisorbashweb fetchweb search
#3
$5.40/1M · $0.0600 per task
The anthropic/claude-sonnet-4.6 model offers robust performance for iterative development and complex codebase navigation at a blended price of $5.40/1M tokens. It is well suited for extensive project management and professional tasks, though budget-conscious operators may find it hard to justify the cost for basic homelab automations.
premiumopen weightslong context9 providersadvisorbashweb search
Models that came with an agent harness bolted on. Point them at a repo and walk away.
#1
$0.168/1M · $0.00196 per task
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting cost of $0.168/1M blended tokens. It is well-suited for processing large codebases and documents in a homelab, though its lower popularity score of 46 means community integration support may be lacking. It is not worth the hardware overhead if your local setup cannot efficiently handle its 309B total parameter mixture-of-experts footprint.
cheapopen weightslong context10 providers
#2
$0.220/1M · $0.00250 per task
Z.ai: GLM 5.3 Flash delivers strong efficiency for coding and long-horizon agent tasks with an arena rating of 1472 and a 1.0M context window. At $0.220/1M blended tokens, it is a capable choice for homelab operators who need a native multimodal model backed by widespread provider availability. However, it may not be worth the cost if your local tasks do not require multimodal capabilities or long-context reasoning.
cheapopen weightslong context38 providers
#3
$0.343/1M · $0.00359 per task
Z.ai: GLM 5.3 is built for complex software engineering and long-horizon agent tasks, boasting an arena rating of 1474 and a popularity score of 58 across a 1.0M context window. At $0.343/1M blended tokens, it is the most expensive option in this lineup, making it worth the money only when you need deep reasoning capabilities for intricate codebases. If your homelab workloads consist of simple scripts or routine automation, the higher cost is unnecessary.
cheapopen weightslong context48 providers
Speech in, speech out, plus the music and clip generators that bill per song.
#1
$0.08 per song
google/lyria-3-pro-preview is part of Google's family of music generation models available through the Gemini API and is well-suited for generating full-length songs at a fixed cost of $0.08 per song. Since it is currently unrated in the arena and has a low popularity score of 4, it may not be worth your money if you need a proven, highly community-vetted production tool rather than a specialized experimental music generator.
cheaplong contextaudio out
#2
$0.04 per clip
google/lyria-3-clip-preview offers a specialized option for creating 30-second duration music clips through the Gemini API at $0.04 per clip. Given its unrated arena status and a low popularity score of 4, it is only worth trying if you specifically need short audio snippets, and it is not worth the money if you require comprehensive audio generation capabilities or widespread community validation.
cheaplong contextaudio out
#3
$4.00/1M · $0.0450 per task
openai/gpt-audio is OpenAI's first generally available audio model, featuring an upgraded decoder designed for more natural-sounding voices and improved voice consistency across a 128K context window. Priced at $4.00 per 1M blended tokens, it is a viable choice for native voice workflows, but might not be worth the money if you only need text-based processing or if its unrated arena performance fails to justify the per-token cost for your homelab use cases.
mid-priceopen weightsaudio out
Open weights (self-hostable) #open-weights
Weights you can download and run on your own box. Costs you GPU time, not tokens.
#1
$0.168/1M · $0.00196 per task
Xiaomi: MiMo-V2.6-Flash offers a massive 1.1M context window and solid performance for an arena rating of 1452 while maintaining a low self-hosting cost of $0.168/1M blended tokens. It is well-suited for processing large codebases and documents in a homelab, though its lower popularity score of 46 means community integration support may be lacking. It is not worth the hardware overhead if your local setup cannot efficiently handle its 309B total parameter mixture-of-experts footprint.
cheapopen weightslong context10 providers
#2
$0.220/1M · $0.00250 per task
Z.ai: GLM 5.3 Flash delivers strong efficiency for coding and long-horizon agent tasks with an arena rating of 1472 and a 1.0M context window. At $0.220/1M blended tokens, it is a capable choice for homelab operators who need a native multimodal model backed by widespread provider availability. However, it may not be worth the cost if your local tasks do not require multimodal capabilities or long-context reasoning.
cheapopen weightslong context38 providers
#3
$0.105/1M · $0.00120 per task
inclusionAI: Ling 3.1 Flash provides an economical self-hosted option at just $0.105/1M blended tokens with a 262K context window. It is worth considering if you need a large mixture-of-experts architecture for budget-constrained local deployments, though it currently lacks an arena rating and holds a low popularity score of 7. It is not worth deploying if your workflows demand proven benchmark performance or robust community backing.
cheapopen weights
Priced per generation, not per token. A different kind of budget.
#1
$0.500/1M · $0.00550 per task
Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) is an extremely cost-efficient option at $0.500/1M blended tokens, making it great for high-velocity developer pipelines and rapid-fire visual exploration. However, with an arena rating of 1313, it may not be worth the money if your homelab tasks require top-tier visual fidelity.
mid-priceimage out
#2
$2.70/1M · $0.0300 per task
Google: Nano Banana 2.1 strikes a solid balance for image generation and editing tasks on the Flash tier, featuring an arena rating of 1426 and a moderate price of $2.70/1M blended tokens. It might not be worth the money for casual users who can get by with a cheaper, faster pipeline for basic visual tasks.
mid-priceweb searchimage out
#3
$9.40/1M · $0.1100 per task
OpenAI: GPT-5.4 Image 2 delivers top-tier performance with an arena rating of 1461 and rich multimodal workflows combining reasoning, coding, and image generation. At $9.40/1M blended tokens, it is definitely not worth the money for simple or low-volume homelab tasks where cheaper models suffice.
premiumopen weightsweb searchimage out
New this week #new
Models that appeared on OpenRouter since the last issue.
Price changes since the last issue, biggest move first.
No price moved more than 5%.
What people are talking about #news
Community and vendor chatter from the last seven days.
Today's papers from the HuggingFace daily feed.
How this was made #method
Sources
- OpenRouter377 models
- LMArena8 boards, 253 unmatched
- Artificial Analysiskey present
- HuggingFace10 papers
- Hacker News110 items
- Redditdisabled (blocked)
- Vendor blogs7 feeds
Price tiers. cheap under $0.50/1M blended tokens, medium $0.50-$5.00, expensive above. Blended is 80% input, 20% output, priced on the cheapest provider.
Value. value = quality / cost of a 10000 token in + 2000 token out task, with a $0.00 cost floor so free models score high without dividing by zero.
Popularity. A blend of arena votes, how many providers serve the model,
HuggingFace trending and how often the model is named in the week's chatter.
The prose is written by google/gemini-3.5-flash-lite, running on the numbers above it. It is told to describe only the models and prices in its input; blocks that mention an unknown model are dropped before publishing. Without an API key the report falls back to templated text and every number still stands.
Every issue so far. One per week, no spam.
-
2026-W43
router: Xiaomi: MiMo-V2.6-Flash · workhorse: Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) · architect: OpenAI: GPT-5.4 · coding agents: Xiaomi: MiMo-V2.6-Flash · tool use: OpenAI: GPT-5.6 Luna · realtime audio: Google: Lyria 3 Pro Preview · open weights: Xiaomi: MiMo-V2.6-Flash · image video: Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)
2026-10-24