gemma3:1b
The smallest open-weight model in the Gemma 3 series released by Google, with approximately one billion parameters. It is built on technology related to Gemini, uses a decoder-only Transformer architecture, and supports offline operation and CPU inference.
Timeline (last 90 days)
The gemma3:1b model is 815MB in size, gemma3:4b is 3.3GB, and gemma3:1b is 15% faster.
On long clauses gemma3:1b stops with incomplete output, for example returning only one key when five were requested and ending with done_reason: stop after 11 tokens.
Running on CPU on the author's machine, explaining a payment order end to end takes about 40 seconds; the author stresses that this is a single measurement on a single machine, and results will differ with different hardware.
A developer wrote a local tool in about 306 lines of Python 3.10 code (using only the standard library) that communicates over HTTP with Ollama on localhost, with language processing performed by the gemma3:1b model, to explain Swiss administrative letters.
gemma3:1b is the smallest member of Google's Gemma 3 series, with about 1 billion parameters, is an open-weights model that can be downloaded and run offline; the Gemma series is built on technology from the same source as Gemini, uses a decoder-only Transformer architecture, and was trained for multilingual capability and instruction following.
On a GPU-less Linux machine (8 vCPU), gemma3:1b takes about 6 to 9 seconds per analysis but often classifies all conversations as asking_questions and gives generic replies; gemma3:4b takes about 15 to 24 seconds per analysis, can select a reasonable stage and reuse facts the seller mentioned, and can reply in Chinese.
After keyword hints were introduced, the 1B model went from always answering asking_questions to usually selecting the correct stage, while the 4B model sometimes disagreed with the hints, which the author believes indicates the model is making its own judgment.
TerraGemma primarily uses the gemma3:1b model, which is approximately 815 MB and completes inference in approximately 500 ms on an ordinary laptop CPU.
All Facts (8)
The gemma3:1b model is 815MB in size, gemma3:4b is 3.3GB, and gemma3:1b is 15% faster.
50%UnverifiedOn long clauses gemma3:1b stops with incomplete output, for example returning only one key when five were requested and ending with done_reason: stop after 11 tokens.
50%UnverifiedRunning on CPU on the author's machine, explaining a payment order end to end takes about 40 seconds; the author stresses that this is a single measurement on a single machine, and results will differ with different hardware.
50%UnverifiedA developer wrote a local tool in about 306 lines of Python 3.10 code (using only the standard library) that communicates over HTTP with Ollama on localhost, with language processing performed by the gemma3:1b model, to explain Swiss administrative letters.
50%Unverifiedgemma3:1b is the smallest member of Google's Gemma 3 series, with about 1 billion parameters, is an open-weights model that can be downloaded and run offline; the Gemma series is built on technology from the same source as Gemini, uses a decoder-only Transformer architecture, and was trained for multilingual capability and instruction following.
50%UnverifiedOn a GPU-less Linux machine (8 vCPU), gemma3:1b takes about 6 to 9 seconds per analysis but often classifies all conversations as asking_questions and gives generic replies; gemma3:4b takes about 15 to 24 seconds per analysis, can select a reasonable stage and reuse facts the seller mentioned, and can reply in Chinese.
50%UnverifiedAfter keyword hints were introduced, the 1B model went from always answering asking_questions to usually selecting the correct stage, while the 4B model sometimes disagreed with the hints, which the author believes indicates the model is making its own judgment.
50%UnverifiedTerraGemma primarily uses the gemma3:1b model, which is approximately 815 MB and completes inference in approximately 500 ms on an ordinary laptop CPU.
50%