AI model
Google DeepMind's Gemma 4, a family of open-weight language models in sizes from small e2b/e4b edge variants to 26B and 31B, usable locally and on-device.
Microsoft's $4,300 Surface Laptop Ultra ran local models generally a bit faster than an M4 Max Mac Studio, The New Stack's reviewer found. The edge was measured on the charger with models sized to the machine, so buyers should check their own model mix.
Reality
Anthropic ran Gemma4-26B-A4B across 15 off-brand Android TV boxes at about $55 each to test whether an AI worm could host itself on IoT gear. Its researchers theorize a working worm could follow by year end. Unmanaged devices need an owner now.
Perspective Coverage
5 publishersReality
Google DeepMind's EmbeddingGemma 2 maps text, images, video and audio into one vector space using 740M Apache-licensed parameters. It is built to run on phones and laptops, so the case for a hosted embedding API now depends on whether its quality holds up on a buyer's own data.
Reality
Google DeepMind released EmbeddingGemma 2, a 740M-parameter open model that runs on a phone and embeds text, code, images, video and audio in one space. Teams running a separate embedder per modality can consolidate on it if Google's reported benchmark numbers hold on their own data.
Reality
Google has released AI Edge Foresight, a Granola-style Mac meeting note-taker that works fully offline on its on-device EmbeddingGemma 2 model. The Verge reported it is free and experimental, so teams barred from cloud transcription can pilot it cheaply.
Perspective Coverage
3 publishersReality
Google released EmbeddingGemma 2, an Apache 2.0 model that maps text, code, images, video and audio into one embedding space with 740M parameters. Phone apps get offline cross-media search from one set of weights, within limits set by a shared 8K-token window and lossy vector truncation.
Perspective Coverage
17 publishersReality
Ollama 0.35.1 picks Gemma 4's prompt template from the model's name, so a copied or renamed 12B model loses four prompt tokens when thinking is off. Teams that save the library model under their own name can pin RENDERER gemma4-large in the Modelfile to keep the template that matched Google's.
Reality
Aleph Alpha released Kolibri, an Apache 2.0 German-English model that activates 3.46B of its 78.1B parameters per token. Each token costs about as much compute as a small model, yet a team hosting it in Europe still has to fit every expert in memory.
Perspective Coverage
6 publishersReality
Ollama since 0.34.4 lets Gemma 4 skip the requested JSON schema, returning bare text with HTTP 200 in 8 of 24 test calls. Until the open fix ships in a release, structured output on local thinking models needs a shape check in the client.
Reality
Repacking Google's quantization-aware-trained Gemma 4 weights without re-rounding fits the 12B model on one TPU v5e chip at bf16-level accuracy. Adopting it means carrying patches to vLLM's TPU backend and working inside a 9,728-token KV cache.
Reality
Google's Cloud Run instances, priced from $5.70 a month, kept an open-source AI agent running with its memory intact in a developer's Preview test. For an agent that mostly waits, it trades a VM's patching for limits the code must be written around.
Reality
Kaggle benchmark results posted on dev.to report that most of 30 vision models reading 336 synthetic Grafana-style panels found the peak but misread the clock. Copilot incident timelines drafted from screenshots need their start times checked by hand.
Reality
SageMaker costs 1.40x a plain EC2 instance per hour to serve one Gemma 4 vLLM build on the same T4 or L4 GPU, a benchmark on dev.to finds. With decode speed matched within 2%, the extra 40% goes to the managed layer around the GPU.
Reality
Gemma 4 decodes on SageMaker's smallest GPU, a T4, at 0.8x an L4's speed with matching outputs from a patched vLLM image, a dev.to benchmark reports. The T4 is the cheaper choice per token only when it rents for under 80% of the L4's hourly rate.
Reality
Repacking Gemma 4's QAT weights with 4-bit embedding tables made decode up to 1.39x faster on one SageMaker L4, according to a dev.to benchmark series. For teams serving Gemma 4 on vLLM, how the weights are stored becomes a setting to measure alongside model size.
Reality
Developer xbill9 rebuilt Google's QAT Gemma 4 26B as int4 and fit it on one TPU v6e with 53,888 tokens of KV cache, against 3,456 for RedHat's FP8 build. Throughput nearly doubles, with accuracy checked on one classification suite.
Reality
Gemma 4 E2B's 4-bit QAT checkpoint decodes 2.05x faster than bf16 on one SageMaker L4, according to a dev.to benchmark. The swap also frees 18% of GPU memory for a 20% larger KV cache, though it runs only through the vLLM container because JumpStart lists no QAT build.
Reality
A single AMD Developer Cloud droplet at $1.99 an hour was inventoried entirely over twelve tag-scoped MCP tools, and two of the readings came back wrong until the output was parsed instead of a status code.
Reality
NVIDIA's SWE-Serve scores the same 627 patches twice on 19 SGLang tasks, once with the live-serving tests and once without. The pass rate falls from 69.4% to 45.9%, and 242 of the 276 live tests came from SGLang itself.
Reality
Red Hat clocks the same 20-call agent task at roughly 45 seconds on a slow backend and about 13 on a fast one. Model choice for agents is turning into a per-call latency budget, with capability as one input.
Reality
Product · August 21, 2026 · One report1 publisher
Build · August 19, 2026 · Confirmed2 publishers