AI Model Radar 2026-W43 · 2026-10-22

Which models should you actually use?

This week saw new open-source tooling emerge with Respan launching its Rust AI gateway built around plugins alongside VoidAccess for dark web OSINT graph generation, while the broader ecosystem navigated infrastructure risks following a major Google Cloud outage affecting European banking operations. Operators should review their self-hosted agent security and access control setups in light of new launches like Alter for multi-user agent access control, and monitor infrastructure resilience against cloud provider dependencies.

issue 2026-W43 generated 2026-10-22 7 new models

The short answer #quick

One pick per job, the highest value model we found that is also popular enough to be worth trusting.

JobModelPriceWhy
Cheap & small (task routing) Z.ai: GLM 5.3 Flash z-ai/glm-5.3-flash $0.044/1M z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M b…
Medium (real work & coding) Qwen: Qwen3.7 Plus qwen/qwen3.7-plus $0.512/1M qwen/qwen3.7-plus offers a massive 1M context window and solid performance for its $0.512/1M blended token price. It is well-suited for processing lar…
Expensive (architecture) OpenAI: GPT-5.4 openai/gpt-5.4 $5.00/1M openai/gpt-5.4 is OpenAI's latest frontier model that unifies the Codex and GPT lines, featuring a 1.1M context window and solid performance in the ex…
Coding & SWE agents Z.ai: GLM 5.3 Flash z-ai/glm-5.3-flash $0.044/1M z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M b…
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, making it well-suited for high-volume, latency-sensitive homelab …
Voice & audio Google: Lyria 3 Pro Preview google/lyria-3-pro-preview $0.08 per song google/lyria-3-pro-preview is well-suited for generating full-length songs through the Gemini API at $0.08 per song. However, it is not worth the mone…
Open weights (self-hostable) Z.ai: GLM 5.3 Flash z-ai/glm-5.3-flash $0.044/1M z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M b…
Image & video Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) google/gemini-3.1-flash-lite-image $0.500/1M google/gemini-3.1-flash-lite-image is Google's fastest and most cost-efficient image model, priced at just $0.500/1M blended tokens with an arena rati…

Cheap & small (task routing) #router

Cheap enough to call on every step. Routing, classification, extraction, summarising.

#1

Z.ai: GLM 5.3 Flash

$0.044/1M · $0.00045 per task

z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M blended tokens, it is not worth skipping if you need a budget-friendly self-hostable option with high popularity and a 1.0M context.

Cheap cheapopen weightslong context35 providers
Quality 1471
Value 100
Popular 36.1k votes
#2

Xiaomi: MiMo-V2.6-Flash

$0.168/1M · $0.00196 per task

xiaomi/mimo-v2.6-flash provides a Mixture-of-Experts architecture with 309B total parameters for demanding open-source workloads. With a price of $0.168/1M blended tokens, it is not worth the money if you prefer lower-cost alternatives unless you specifically require its 1.1M context capacity.

Cheap cheapopen weightslong context10 providers
Quality 1453
Value 99
Popular 12.6k votes
#3

Google: Gemma 4 31B

$0.240/1M · $0.00280 per task

google/gemma-4-31b-it delivers a dense multimodal experience featuring a 262K context window and a configurable thinking mode. Priced at $0.240/1M blended tokens, it is not worth deploying if you are strictly optimizing for the lowest inference cost among available open weights options.

Cheap cheapopen weights15 providers
Quality 1425
Value 69
Popular 12.3k votes

Medium (real work & coding) #workhorse

The models that do the actual work: writing, refactoring, debugging, agentic loops.

#1

Qwen: Qwen3.7 Plus

$0.512/1M · $0.00576 per task

qwen/qwen3.7-plus offers a massive 1M context window and solid performance for its $0.512/1M blended token price. It is well-suited for processing large codebases and long documents in homelab automation pipelines without breaking the bank. However, with an arena rating of 1454, it may not be worth the money if you require absolute peak accuracy for complex programming tasks.

Medium mid-pricelong context
Quality 1454
Value 100
Popular 42.0k votes
#2

Xiaomi: MiMo-V2.6-Pro

$0.522/1M · $0.00609 per task

xiaomi/mimo-v2.6-pro stands out as the strongest performer in this tier with an arena rating of 1491 and a massive 1.1M context window for $0.522/1M blended tokens. It is an excellent choice for demanding real-world coding projects and heavy technical workloads where capability is paramount. Given its low popularity score of 43, it is an under-the-radar option that punches well above its weight class for the price.

Medium mid-priceopen weightslong context5 providers
Quality 1491
Value 97
Popular 4.1k votes
#3

Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)

$0.500/1M · $0.00550 per task

google/gemini-3.1-flash-lite-image is Google's fastest and most cost-efficient image model, priced at just $0.500/1M blended tokens with an arena rating of 1313. It is ideal for high-velocity developer pipelines and rapid-fire visual exploration where speed matters. However, it is not worth the money if you require top-tier aesthetic depth or complex multimodal reasoning.

Medium mid-priceimage out
Quality 1313
Value 95
Popular 65.7k votes

Expensive (architecture) #architect

When the answer has to be right and the reasoning has to hold up. Use sparingly.

#1

OpenAI: GPT-5.4

$5.00/1M · $0.0550 per task

openai/gpt-5.4 is OpenAI's latest frontier model that unifies the Codex and GPT lines, featuring a 1.1M context window and solid performance in the expensive category. It is well suited for demanding tasks requiring a massive context window and top-tier arena ratings. However, given its high blended cost of $5.00/1M tokens, it is not worth the money for casual homelab workloads or simple automation that cheaper models can handle.

Expensive premiumopen weightslong context7 providersapply patchweb search
Quality 1469
Value 100
Popular 126.7k votes
#2

Anthropic: Claude Sonnet 4.6

$5.40/1M · $0.0600 per task

anthropic/claude-sonnet-4.6 is Anthropic's most capable Sonnet-class model, offering frontier performance across coding, agents, and professional work with a 1M context window. It is ideal for iterative software development, complex codebase navigation, and end-to-end project management. With a price of $5.40/1M blended tokens, it is not worth the money if your homelab tasks do not require advanced agentic workflows or heavy coding capabilities.

Expensive premiumopen weightslong context9 providersadvisorbashweb search
Quality 1458
Value 91
Popular 174.0k votes
#3

Anthropic: Claude Sonnet 4.5

$5.40/1M · $0.0600 per task

anthropic/claude-sonnet-4.5 is an advanced Sonnet model optimized for real-world agents and coding workflows, delivering strong benchmark performance and a 1M context window. It excels at tasks that demand specialized coding logic and reliable agent-based execution. At $5.40/1M blended tokens, it is not worth the money since the newer version provides higher quality at a similar price point.

Expensive premiumopen weightslong context7 providersadvisorbashweb search
Quality 1450
Value 90
Popular 31.5k votes

Coding & SWE agents #coding-agents

Models that came with an agent harness bolted on. Point them at a repo and walk away.

#1

Z.ai: GLM 5.3 Flash

$0.044/1M · $0.00045 per task

z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M blended tokens, it is not worth skipping if you need a budget-friendly self-hostable option with high popularity and a 1.0M context.

Cheap cheapopen weightslong context35 providers
Quality 1471
Value 100
Popular 36.1k votes
#2

Xiaomi: MiMo-V2.6-Flash

$0.168/1M · $0.00196 per task

xiaomi/mimo-v2.6-flash provides a Mixture-of-Experts architecture with 309B total parameters for demanding open-source workloads. With a price of $0.168/1M blended tokens, it is not worth the money if you prefer lower-cost alternatives unless you specifically require its 1.1M context capacity.

Cheap cheapopen weightslong context10 providers
Quality 1453
Value 99
Popular 12.6k votes
#3

Xiaomi: MiMo-V2.5-Pro

$0.365/1M · $0.00426 per task

xiaomi/mimo-v2.5-pro serves as Xiaomi's flagship model, delivering strong performance in general agentic capabilities, complex software engineering, and long-horizon tasks with top rankings on benchmarks like SWE-bench Pro. At $0.365 per 1M blended tokens and an arena rating of 1464, it provides the advanced reasoning required for demanding software development workloads. It is not worth the money, though, if your tasks are straightforward enough to be handled reliably by less expensive flash-tier alternatives.

Cheap cheapopen weightslong context4 providers
Quality 1464
Value 47
Popular 77.4k votes

Out of the box (native tools) #tool-use

Native web search, bash and code interpreter. The tool calls are not a prompt template.

#1

OpenAI: GPT-5.6 Luna

$0.400/1M · $0.00440 per task

openai/gpt-5.6-luna is a fast and cost-efficient option at $0.400/1M blended tokens, making it well-suited for high-volume, latency-sensitive homelab tasks like chat, classification, and lightweight agentic workflows. However, it may not be worth the money if your use case demands deep, multi-step reasoning where its lower arena rating and simpler capabilities fall short.

Cheap cheaplong context7 providersapply patchweb search
Quality 1457
Value 100
Popular 5.8k votes
#2

Google: Gemini 3 Flash Preview

$1.00/1M · $0.0110 per task

google/gemini-3-flash-preview provides near Pro-level reasoning and tool use at a blended price of $1.00/1M tokens, making it a high-value choice for multi-turn chat, coding assistance, and agentic workflows. It is not worth the money if you require the absolute highest benchmark performance, since its predecessor positioning offers slightly lower overall capabilities than its full-line sibling.

Medium mid-priceopen weightslong context6 providersweb search
Quality 1466
Value 40
Popular 196.4k votes
#3

Google: Gemini 3.7 Flash

$1.35/1M · $0.0150 per task

google/gemini-3.7-flash is designed for fast agentic workflows, coding, and complex multi-step reasoning, boasting the highest arena rating in this lineup at 1489. At $1.35/1M blended tokens, it is not worth the money for simple, high-volume tasks where a cheaper alternative can handle the workload adequately without paying a premium for its advanced multi-step capabilities.

Medium mid-pricelong context7 providersweb search
Quality 1489
Value 30
Popular 29.2k votes

Voice & audio #realtime-audio

Speech in, speech out, plus the music and clip generators that bill per song.

#1

Google: Lyria 3 Pro Preview

$0.08 per song

google/lyria-3-pro-preview is well-suited for generating full-length songs through the Gemini API at $0.08 per song. However, it is not worth the money if you only need short background cues or quick musical loops, as the clip version is cheaper for that purpose.

Cheap cheaplong contextaudio out
Quality unrated
Value 0
Popular -
#2

Google: Lyria 3 Clip Preview

$0.04 per clip

google/lyria-3-clip-preview excels at producing 30-second duration music clips at a lower cost of $0.04 per clip via the Gemini API. It is not worth the money if you require complete, full-length tracks, which necessitate the pro variant instead.

Cheap cheaplong contextaudio out
Quality unrated
Value 0
Popular -
#3

OpenAI: GPT Audio

$4.00/1M · $0.0450 per task

openai/gpt-audio is a solid choice for applications needing natural-sounding voices and reliable voice consistency with its upgraded decoder. At $4.00 per 1M blended tokens, it may not be worth the money for casual homelab workloads that do not strictly require advanced audio generation and decoding capabilities.

Medium mid-priceopen weightsaudio out
Quality unrated
Value 0
Popular -

Open weights (self-hostable) #open-weights

Weights you can download and run on your own box. Costs you GPU time, not tokens.

#1

Z.ai: GLM 5.3 Flash

$0.044/1M · $0.00045 per task

z-ai/glm-5.3-flash is well suited for efficient coding and long-horizon agent tasks while handling extended context windows affordably. At $0.044/1M blended tokens, it is not worth skipping if you need a budget-friendly self-hostable option with high popularity and a 1.0M context.

Cheap cheapopen weightslong context35 providers
Quality 1471
Value 100
Popular 36.1k votes
#2

Xiaomi: MiMo-V2.6-Flash

$0.168/1M · $0.00196 per task

xiaomi/mimo-v2.6-flash provides a Mixture-of-Experts architecture with 309B total parameters for demanding open-source workloads. With a price of $0.168/1M blended tokens, it is not worth the money if you prefer lower-cost alternatives unless you specifically require its 1.1M context capacity.

Cheap cheapopen weightslong context10 providers
Quality 1453
Value 99
Popular 12.6k votes
#3

Google: Gemma 4 31B

$0.240/1M · $0.00280 per task

google/gemma-4-31b-it delivers a dense multimodal experience featuring a 262K context window and a configurable thinking mode. Priced at $0.240/1M blended tokens, it is not worth deploying if you are strictly optimizing for the lowest inference cost among available open weights options.

Cheap cheapopen weights15 providers
Quality 1425
Value 69
Popular 12.3k votes

Image & video #image-video

Priced per generation, not per token. A different kind of budget.

#1

Google: Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)

$0.500/1M · $0.00550 per task

google/gemini-3.1-flash-lite-image is Google's fastest and most cost-efficient image model, priced at just $0.500/1M blended tokens with an arena rating of 1313. It is ideal for high-velocity developer pipelines and rapid-fire visual exploration where speed matters. However, it is not worth the money if you require top-tier aesthetic depth or complex multimodal reasoning.

Medium mid-priceimage out
Quality 1313
Value 100
Popular 65.7k votes
#2

Google: Nano Banana 2.1

$2.70/1M · $0.0300 per task

google/gemini-nano-banana-2.1 delivers solid image generation and editing capabilities on the Flash tier, achieving an arena rating of 1426 for $2.70/1M blended tokens. It excels at tasks like product recontextualization in automated workflows. On the other hand, it might not be worth the money if you need the absolute highest fidelity or advanced multi-step reasoning found in pricier alternatives.

Medium mid-priceweb searchimage out
Quality 1426
Value 20
Popular 12.9k votes
#3

OpenAI: GPT-5.4 Image 2

$9.40/1M · $0.1100 per task

openai/gpt-5.4-image-2 combines OpenAI's GPT-5.4 reasoning with advanced image generation to power rich multimodal workflows, hitting the highest arena rating of 1461. It is worth the money for complex tasks that require seamless transitions between coding, deep reasoning, and high-end visual tasks. However, at $9.40/1M blended tokens, it is overkill and not worth the money for simple or standalone image generation pipelines.

Expensive premiumopen weightsweb searchimage out
Quality 1461
Value 6
Popular 328.3k votes

New this week #new

Models that appeared on OpenRouter since the last issue.

Price watch #prices

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.

Reading #papers

Today's papers from the HuggingFace daily feed.

How this was made #method

Sources

  • OpenRouter0 models
  • LMArena8 boards, 252 unmatched
  • Artificial Analysiskey present
  • HuggingFace10 papers
  • Hacker News111 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.

Archive #archive

Every issue so far. One per week, no spam.