• Latest
  • Trending
  • All
Official Apple product photograph of the 2026 Mac mini, a small square silver aluminium desktop computer with rounded corners and a dark circular base, shown at a three quarter angle against a plain light background.

The M6 Mac mini is $899, capped at 16GB

3 September 2026
Answer card stating that Qwen-Image-2.1, released on 20 September 2026, ships open weights with a 7 billion parameter diffusion transformer, a Qwen3-VL 8B text encoder and an RGBA VAE totalling about 33 gigabytes in BF16, under the Qwen Research License that limits use to research or evaluation and requires a separate commercial licence, unlike the Apache 2.0 licence of Qwen-Image 1.0.

Qwen-Image-2.1 brings the weights back, but not the Apache licence

21 September 2026
Answer card stating that Ternary Bonsai 2 27B, released by PrismML on 17 September 2026 under Apache 2.0, packs Qwen3.8 27B into 5.95 gigabytes at 1.72 bits per weight, keeps 98.2 percent of the 14-benchmark average, about 75 percent on SWE-bench Verified and Terminal-Bench 2.1, and needs PrismML's llama.cpp fork to run.

Does Bonsai 2 27B really keep 98% of Qwen3.8 in 5.95 GB?

20 September 2026
Answer card stating that Jev 1.13 from TypeSafe AI is a decision model in early access since 15 September 2026 that returns typed probabilities instead of text, priced at 42 dollars per billion input tokens with output tokens free, answering in 70 to 500 milliseconds, with a 64K token request budget, text input only, and a documented list of things it does badly, including counting and dates.

Jev 1.13 bills $42 a billion tokens, and it can’t count

19 September 2026
Answer card stating that Qwen3.8-Omni-Flash launched on 17 September 2026 as an API only model on Alibaba Cloud Model Studio, taking text, images, audio and video in a 1M token context and returning text only, priced at 0.15 dollars per million input tokens for every modality and 0.47 dollars per million output tokens in the international regions, with no open weights published and the Qwen-Live Harness GitHub repository returning 404.

Qwen3.8-Omni-Flash bills audio at $0.15 and ships no weights

18 September 2026
Answer card stating that on 15 September 2026 AWS said it is unable to restore access to resources and data hosted exclusively in the Middle East Bahrain region me-south-1 and in the mec1-az2 zone of the UAE region, because the damage spanned multiple Availability Zones and exceeded what multi-AZ services are designed to withstand.

AWS can’t restore me-south-1, six months after the drone strikes

17 September 2026
Answer card stating that Google released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking on 15 September 2026 at 3 dollars per million audio input tokens and 12 dollars out, that the thinking model requires asynchronous tools, and that Artificial Analysis scores it 82.6 on its Speech to Speech Quality Index.

Gemini 3.8 Live Extended Thinking rejects any tool that blocks

16 September 2026
Answer card summarising the Atria Dawn Preview release: 744B GLM-5.2 base, MIT licence, 1.5 TB BF16 and 756 GB FP8 checkpoints, 256K context, top on five of sixteen benchmark rows and trailing on SWE-bench Pro.

Atria Dawn Preview is 744B under MIT, and the BF16 weighs 1.5 TB

15 September 2026
Answer card stating that OpenAI released the Agents API in public beta on 10 September 2026 with no separate fee, billed through model tokens, tool calls and hosted sandbox time, with a choice of OpenAI hosted, self hosted or partner sandboxes, US only data residency and no Zero Data Retention support.

OpenAI’s Agents API has no fee, no ZDR and a one hour sandbox clock

14 September 2026
Answer card: Sakana Fugu Max at $2 and $6 per million tokens, Fugu Ultra v2 unchanged at $5 and $30, and Sakana saying Ultra v2 scores without Fable 5 or GPT-6 Astra in its pool.

Fugu Max costs $2 and $6 while Fugu Ultra v2 runs without Fable 5

13 September 2026
Answer card stating that DeepSeek released DeepSeek-V4.1-Flash on 10 September 2026 as a 552 billion parameter mixture of experts model with a new causal encoder decoder architecture that activates 8 billion parameters on input and 16 billion on output, with native vision, a one million token context and MIT licensed weights, that the API model name is now deepseek-flash at 0.15 dollars per million input tokens and 0.60 dollars per million output tokens off peak, and that DeepSeek announced V4 Pro would be routed to V4.1-Flash from 14 September and reversed that on 11 September.

DeepSeek V4.1-Flash arrived, and the V4 Pro retirement lasted a day

12 September 2026
Answer card stating that Cognition released SWE-2 on 10 September 2026, a coding model post-trained from Kimi K3, scoring 50.0 percent on FrontierCode 1.1 Main against 50.9 percent for Claude Fable 5.1 and 27.3 percent on Terminal-Bench 4 against 55.8 percent, available only inside Devin.

SWE-2 trails Fable 5.1 by one point, and by 28 on Terminal-Bench 4

11 September 2026
Answer card for Meta Muse, free to 100 million tokens a week then $20 a month, launched 8 September 2026 for United States adults only, running in a dedicated per user virtual machine.

Does Meta Muse do enough to earn your inbox and a card on file?

9 September 2026
  • About
  • Contact
  • Privacy
  • Legal
Tuesday, September 22, 2026
  • Login
Packet Nebula
  • Home
  • Articles
    • Security
    • Network
    • Dev
    • Sysadmin
    • SEO
    • Email & DNS
  • Tools
    • Network tools: free, fast, no signup
    • Security tools: free, fast, no signup
    • Developer tools: free, fast, no signup
    • Sysadmin tools: free, fast, no signup
    • SEO tools: free, fast, no signup
    • Email & DNS tools: free, fast, no signup
  • Download
  • About
No Result
View All Result
Packet Nebula
No Result
View All Result
Home Dev

The M6 Mac mini is $899, capped at 16GB

by stephane
3 September 2026
in Dev
0
Official Apple product photograph of the 2026 Mac mini, a small square silver aluminium desktop computer with rounded corners and a dark circular base, shown at a three quarter angle against a plain light background.
500
SHARES
1.4k
VIEWS
Share on FacebookShare on Twitter

The cheap Mac is gone. Apple opened pre-orders on the M6 Mac mini on 25 August, and the entry machine now costs $899 with 16GB of unified memory and a 256GB SSD, which is the exact configuration that sat at $599 when the M4 model launched in November 2024. Same memory, same storage, three hundred dollars more. You do get real hardware for it: a 12-core CPU, a 12-core GPU with Neural Accelerators in every core, a dual 16-core Neural Engine, 170GB/s of bandwidth, and 2.5Gb Ethernet instead of gigabit. Whether that adds up depends entirely on what you were planning to put on the box, and if the answer is local models, read the memory line twice.

The short answer

Apple opened pre-orders for the M6 Mac mini on 25 August 2026, shipping 22 September. The M6 brings a 12-core CPU, a 12-core GPU with Neural Accelerators in each core, a dual 16-core Neural Engine and 170GB/s of bandwidth. Wi-Fi 7, Bluetooth 6 and 2.5Gb Ethernet across the range. The catch is the price and the memory: $899 buys the same 16GB and 256GB that cost $599 at the M4 launch, and 32GB is as far as the M6 goes. The M5 Pro at $1,699 gets 24GB standard, up to 64GB, 307GB/s and Thunderbolt 5.

$899M6 base, 16GB and 256GB
2.5GbEthernet now standard, 10Gb optional
32GBhard memory ceiling on the M6
Answer card stating that Apple opened pre-orders for the M6 Mac mini on 25 August 2026 at 899 dollars for a 12-core CPU with 16GB of unified memory and a 256GB SSD, shipping 22 September, with the same configuration having cost 599 dollars at the M4 launch in November 2024.
Same memory. Same storage. Three hundred dollars of daylight.

What $899 actually gets you

The chip is a genuine step. M6 runs a 12-core CPU and a 12-core GPU, both up two cores from the M4, and for the first time on a Mac mini every GPU core carries a Neural Accelerator. The Neural Engine is now a dual 16-core arrangement rather than a single block, and memory bandwidth climbs to 170GB/s from the M4’s 153GB/s.

Apple’s headline claims, all measured against the M4 Mac mini: up to 40 percent faster CPU, up to 2x graphics, up to 2x storage, up to 4x on AI work. The one number with an actual application attached is LM Studio prompt processing, quoted at up to 4.8x faster than M4 and 13.5x faster than M1. Vendor benchmarks on vendor workloads. We’d wait for someone independent to run a real model before treating any of it as planning data.

Two things in the spec sheet matter more to us than any of that.

Ethernet finally moves off gigabit. Every new Mac mini ships with 2.5Gb as standard, with the 10Gb option still there at extra cost. For anyone using one of these as an always-on box next to a NAS, that’s the upgrade that changes a working day, and it’s the one Apple buried three quarters of the way down the press release. Wi-Fi 7 and Bluetooth 6 come along too, up from Wi-Fi 6E and Bluetooth 5.3.

Bar chart of the entry price for the base Mac mini in US dollars, showing 599 dollars for the M4 at its November 2024 launch, 799 dollars for the mid 2026 entry configuration with a 512GB SSD, and 899 dollars for the M6 announced on 25 August 2026 with 16GB of memory and a 256GB SSD.
The 256GB entry config came back. At a different price.

The price line, without the spin

Here’s the sequence, and we’re keeping it to things that are on the record.

November 2024: the M4 Mac mini arrives at $599 with 16GB and a 256GB SSD. During 2026 that configuration disappears from the store, and the entry point becomes a $799 machine with a 512GB SSD. In June, the M4 Pro version goes from $1,399 to $1,599. Now the M6 puts 16GB and 256GB back on the shelf at $899.

So the like-for-like comparison is $599 to $899. Fifty percent, in under two years, on the machine whose entire reputation was being the cheap way into a Mac.

Apple gave no reason, and we’re not going to invent one on its behalf. The context everyone is pointing at is real enough: memory and NAND contract prices have been climbing hard all year as data centre buyers absorb supply, and we’ve watched the same pressure land on the MacBook Air and on Pixel pricing. Whether that fully explains a $300 move on a $599 product, I honestly doubt. Component cost is part of it. The rest is what a company charges when the thing sells anyway.

Apple is selling this as an AI box. Read the memory line.

The press release says “always-on agentic computing” and it says it more than once. Srouji’s quote calls it an always-on agentic device. The marketing images show local models and AI agents automating daily tasks. That’s the pitch.

And then the $899 machine ships with 16GB of unified memory, configurable to 32GB, and that is the ceiling.

Unified memory is shared. Whatever the model loads comes out of the same pool as macOS, your browser and everything else running. On 16GB you’re comfortable with quantised weights in roughly the 7B to 14B band plus a usable context window, which genuinely covers a lot of agent work and most coding assistants. Push into the 30B class and you need 32GB, which is the M6’s maximum. Past that you’re buying the M5 Pro, which starts at 24GB, goes to 64GB, and runs 307GB/s of bandwidth against the M6’s 170.

The port split says the same thing in hardware. M6 gets three Thunderbolt 4 ports. M5 Pro gets three Thunderbolt 5. Apple explicitly states that Thunderbolt 5 is what allows clustering several Mac mini systems together to run large AI models entirely on device. That capability simply isn’t on the cheap machine, by design.

Official Apple photograph of the rear of the 2026 Mac mini showing three Thunderbolt ports, an HDMI port, an Ethernet jack and the power connector along the back edge of the silver aluminium chassis.

Image: Apple, from its Mac mini newsroom announcement, 25 August 2026.

I might be wrong about how much this bites. For an always-on agent chewing through an 8B model and a queue of small tasks, 16GB is fine and the extra Neural Engine will show up in latency you can feel. But the gap between the copy and the configuration is wide enough to notice.

Checklist comparing the M6 and M5 Pro Mac mini, showing the M6 at 899 dollars suits an always on desktop with 2.5Gb Ethernet and small local models up to about 14 billion parameters but caps at 32GB of memory with Thunderbolt 4 and no clustering, while the M5 Pro at 1699 dollars offers 24GB standard up to 64GB, 307GB per second of bandwidth and Thunderbolt 5 clustering for larger models.
Two machines with the same shell and very different ceilings.

So which one, and when

If you want a small, quiet, low-draw desktop that runs all the time and talks to a 2.5Gb switch, the M6 at $899 does that well and the chip is not the limiting factor. Buy it for what it is.

If the reason you’re looking at a Mac mini is running models locally, the $899 configuration is the wrong purchase and the 32GB upgrade only moves the wall a little. Either budget for the M5 Pro or wait, because the memory market is the thing setting these prices and nobody sensible is predicting where it lands.

Worth a note on timing: pre-orders opened 25 August in 30 countries, nothing ships until 22 September, and macOS 27 arrives as a free update in the autumn. If your workflow depends on the new Apple Intelligence and Siri work Apple keeps attaching to this launch, that’s a second date to watch, and it isn’t the one on your delivery slip. The previous generation sold out fast enough to show up on eBay above list, so if you’ve decided, decide early.

Sources

Chip specifications, the 170GB/s and 307GB/s bandwidth figures, the port breakdown, the 2.5Gb Ethernet and Wi-Fi 7 details, the Thunderbolt 5 clustering statement, the Srouji quote, the LM Studio comparisons and the pre-order and availability dates all come from Apple’s own Mac mini announcement, published 25 August 2026, including its education pricing note. Base configuration and pricing for both models, plus the discontinuation of the $599 machine, are reported by TechCrunch and MacRumors. The price history, including the mid-cycle move to a $799 entry point and the M4 Pro increase from $1,399 to $1,599 in June, is from Macworld. M4 Mac mini reference specifications are from Apple’s own tech specs page.

Frequently asked questions

How much does the M6 Mac mini cost?

The M6 model starts at $899 in the US with a 12-core CPU, 12-core GPU, 16GB of unified memory and a 256GB SSD, or $799 with education pricing. The M5 Pro model starts at $1,699, or $1,599 education, and comes with 24GB of memory and a 512GB SSD. Pre-orders opened on 25 August 2026 in 30 countries and regions, with deliveries and retail availability starting 22 September.

Why did the Mac mini go from $599 to $899?

Apple has not published a reason, and we would not quote one it did not give. What is on the record is the sequence: the M4 Mac mini launched at $599 in November 2024 with 16GB and 256GB, that configuration was dropped during 2026 leaving a $799 model with a 512GB SSD as the entry point, the M4 Pro version went from $1,399 to $1,599 in June, and the M6 now reinstates 16GB and 256GB at $899. Memory and NAND contract prices have been climbing all year on data centre demand, which is the obvious backdrop, but Apple itself stayed quiet about it.

Is 16GB enough to run local AI models on the M6 Mac mini?

For small models, yes. Unified memory is shared between the system and the model, so on a 16GB machine you realistically fit quantised models in the 7B to 14B range with room for a context window, which covers a lot of agent and coding assistant work. Anything in the 30B and up class needs the 32GB configuration at minimum, and the M6 tops out at 32GB. If you want to load bigger weights, the M5 Pro goes to 64GB with 307GB/s of bandwidth.

What is the difference in ports between the M6 and M5 Pro Mac mini?

Both have two front USB-C ports supporting USB 3, a front headphone jack, HDMI, and Ethernet at 2.5Gb with a 10Gb option. The split is on the back: three Thunderbolt 4 ports on the M6, three Thunderbolt 5 ports on the M5 Pro. Apple says Thunderbolt 5 is what lets you cluster several Mac mini systems to run large AI models entirely on device, so that capability does not exist on the $899 machine.

How much faster is the M6 Mac mini than the M4?

Apple claims up to 40 percent faster CPU, up to 2x graphics, up to 2x storage and up to 4x AI performance versus the M4 Mac mini. The most concrete figure it published is LLM prompt processing in LM Studio, quoted as up to 4.8x faster than M4 and up to 13.5x faster than M1. These are vendor numbers on vendor-chosen workloads, so treat them as a direction rather than a benchmark, and wait for independent runs before budgeting around them.

Tags: aibig-techhardwarellmnewspricing
Share200Tweet125
stephane

stephane

  • Trending
  • Comments
  • Latest
Answer card: Proton Lumo 2.0 is private by policy, not by locality. Saved history is locked so even Proton cannot read it, but the prompt is decrypted on a Proton EU server to answer it, then forgotten.

Proton Lumo 2.0 review: how private is it, really?

3 September 2026
The Agentic Coding section of the official Hy4 preview benchmark appendix published by Tencent, a table comparing Hy3 and Hy4 preview against DeepSeek V4 Pro 0813, Qwen 3.8 Max, GLM 5.3, Kimi K3, GPT 5.6 Sol and Claude Opus 5 across SWE-bench Multilingual, SWE-bench Pro, DeepSWE, three SWE Atlas tasks, SWE-Marathon, Terminal-Bench 2.1, NL2Repo-Bench, CyberGym, ProgramBench, PostTrainBench and Harbor-Index.

Tencent’s 770B Hy4 tops one benchmark row in 46

3 September 2026
Answer card: Qwen 3.7 Max is API-only and cannot run locally yet; the open Qwen models (Qwen 3.6 27B, qwen3:8b to 32b) run offline via Ollama.

Qwen 3.7 local: what you can actually run offline

22 June 2026
Answer card: JWTs are not encrypted, anyone can read them; the signature proves who issued the token, not who may read it.

Are JWTs encrypted? No, and the difference will bite you

0
Answer card: a random 8 character password falls in under 2 hours offline, while 16 random characters hold for 1.4 trillion years at the same speed.

How long does it take to crack a password in 2026?

0
Answer card: three DNS records decide if your mail lands or bounces; SPF lists allowed senders, DKIM signs messages, DMARC sets the failure policy.

SPF, DKIM and DMARC explained: the records your email needs

0
Answer card stating that Qwen-Image-2.1, released on 20 September 2026, ships open weights with a 7 billion parameter diffusion transformer, a Qwen3-VL 8B text encoder and an RGBA VAE totalling about 33 gigabytes in BF16, under the Qwen Research License that limits use to research or evaluation and requires a separate commercial licence, unlike the Apache 2.0 licence of Qwen-Image 1.0.

Qwen-Image-2.1 brings the weights back, but not the Apache licence

21 September 2026
Answer card stating that Ternary Bonsai 2 27B, released by PrismML on 17 September 2026 under Apache 2.0, packs Qwen3.8 27B into 5.95 gigabytes at 1.72 bits per weight, keeps 98.2 percent of the 14-benchmark average, about 75 percent on SWE-bench Verified and Terminal-Bench 2.1, and needs PrismML's llama.cpp fork to run.

Does Bonsai 2 27B really keep 98% of Qwen3.8 in 5.95 GB?

20 September 2026
Answer card stating that Jev 1.13 from TypeSafe AI is a decision model in early access since 15 September 2026 that returns typed probabilities instead of text, priced at 42 dollars per billion input tokens with output tokens free, answering in 70 to 500 milliseconds, with a 64K token request budget, text input only, and a documented list of things it does badly, including counting and dates.

Jev 1.13 bills $42 a billion tokens, and it can’t count

19 September 2026
  • About
  • Contact
  • Privacy
  • Legal

Copyright © 2026 Stephane Cardon.

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • Home
  • Articles
    • Security
    • Network
    • Dev
    • Sysadmin
    • SEO
    • Email & DNS
  • Tools
    • Network tools: free, fast, no signup
    • Security tools: free, fast, no signup
    • Developer tools: free, fast, no signup
    • Sysadmin tools: free, fast, no signup
    • SEO tools: free, fast, no signup
    • Email & DNS tools: free, fast, no signup
  • Download
  • About

Copyright © 2026 Stephane Cardon.