/inferences/gemma-4/e2b/unsloth-q4_k_m-ollama

hf.co/unsloth/gemma-4-E2B-it-GGUF:Q4_K_M

unsloth · Q4_K_M · on ollama · quantizer build

Measured by 8 nodes on hardware we do not own, and signed by each of them. What machines here actually proved — never what the card claims.

matrix built 24 Oct 2026 · every field verified against the signature of the node that produced it, and re-verifiable by you: each signed figure carries the exact bytes the node signed (signing_payload_b64) and the signature over them, and the ed25519 public key is embedded in node_did. effective_ctx is the context length a node PROVED by recall, never the length a model card advertises
context declared1,31,072what the runtime reported. A node attests it was told this; nobody attests it is true.
context proved by recall1,31,072the largest size at which a machine could still find tokens planted across the whole prompt — start, middle and end, all of which had to come back.Also served single-stream at 32,768 — the size a machine loaded and answered at, which is a weaker claim than recall and is shown beside it rather than merged into it.
A machine here made it use everything the runtime declared.
recalled 128k▲ served 32k⌐ declared 128k · unsigned
every machine that timed it
87.4 tok/s · unnamed machine77.6 tok/s · unnamed machine74.4 tok/s · NVIDIA · Washington27.4 tok/s · unnamed machine15.3 tok/s · unnamed machine10.0 tok/s · unnamed machine6.8 tok/s · x86 · CPU · Hyderabad4.0 tok/s · Apple silicon · Bhubaneswar
4.0 → 87.4 tok/s across 8 machines · ×21.9 spreadunnamed machineNVIDIAx86 · CPUApple silicon

This model calls tools correctly and does not stop. A node watched it call the probe tool, then call it again, until the turn budget ran out. Chosen on “supports tools” alone, it will call a tool in production and never come back.

How it was served — proved batch profiles

the same model advertises itself in different shapes, each one a machine proved
077155097194tok/s total — what the machine serves →tok/s per user →4 slots at 32k context · 172.9 tok/s total · 46.9 tok/s per user · long-context4×4 slots at 8k context · 168.7 tok/s total · 46.1 tok/s per user · high-throughput4×4 slots at 8k context · 84.2 tok/s total · 56.3 tok/s per user · high-throughput4×4 slots at 8k context · 29.5 tok/s total · 8.7 tok/s per user · high-throughput4×1 slot at 8k context · 10.7 tok/s total · 75.6 tok/s per user · interactive1×1 slot at 8k context · 10.4 tok/s total · 134.5 tok/s per user · interactive1×1 slot at 32k context · 10.4 tok/s total · 78.6 tok/s per user · long-context1×1 slot at 8k context · 3.2 tok/s total · 11.7 tok/s per user · interactive1×
Each circle is one signed batch profile; the number above it is how many replies were decoding at once, and the circle grows with the context it was proved at. Up and to the right is better and mostly unreachable — the curve bending down as you go right is the trade: more people served, each of them waiting longer.
4 users, batchedlong-context
32kcontext
173tok/s total
46.9tok/s · per user
4 replies decoding at once — the capacity this gram can actually serve a room.
7.0 GiB resident687 MiB KV per user
did:epn:002408…a5be46 · q4_0 KVsigned
4 users, batchedhigh-throughput
8kcontext
169tok/s total
46.1tok/s · per user
4 replies decoding at once — the capacity this gram can actually serve a room.
5.0 GiB resident171 MiB KV per user
did:epn:002408…a5be46 · q4_0 KVsigned
4 users, batchedhigh-throughput
8kcontext
84tok/s total
56.3tok/s · per user
4 replies decoding at once — the capacity this gram can actually serve a room.
5.0 GiB resident171 MiB KV per user
did:epn:002408…84bf54 · q4_0 KVsigned
4 users, batchedhigh-throughput
8kcontext
30tok/s total
8.7tok/s · per user
4 replies decoding at once — the capacity this gram can actually serve a room.
5.0 GiB resident171 MiB KV per user
did:epn:002408…9e3c79 · q4_0 KVsigned
Single streaminteractive
8kcontext
11tok/s total
75.6tok/s · per user
One reply, streaming — the latency a single user sees.
4.5 GiB resident171 MiB KV per user
did:epn:002408…a5be46 · q4_0 KVsigned
Single streaminteractive
8kcontext
10tok/s total
134.5tok/s · per user
One reply, streaming — the latency a single user sees.
4.5 GiB resident171 MiB KV per user
did:epn:002408…84bf54 · q4_0 KVsigned
Single streamlong-context
32kcontext
10tok/s total
78.6tok/s · per user
One reply, streaming — the latency a single user sees.
5.0 GiB resident687 MiB KV per user
did:epn:002408…a5be46 · q4_0 KVsigned
Single streaminteractive
8kcontext
3tok/s total
11.7tok/s · per user
One reply, streaming — the latency a single user sees.
4.5 GiB resident171 MiB KV per user
did:epn:002408…9e3c79 · q4_0 KVsigned

Languages, proved

AmharicArabicAssameseBengaliChineseDutchEnglishFrenchGermanGujaratiHebrewHindiIndonesianItalianJapaneseKannadaKoreanMalayMalayalamMarathiNepaliOdiaPersianPolishPortuguesePunjabiRussianSinhalaSpanishSwahiliTamilTeluguThaiTurkishUkrainianUrduVietnamese

Each question was put entirely in its own language and script, and the answer had to be right. A model with no coverage cannot parse the question, so answering in English is a failure here, not a pass. 37 proved of 39 asked — a language absent from the probe set was never asked, and is not a claim either way.

Capabilities, measured

Audioprobed, and failed
Given audio and asked to transcribe it.
Reasons about codepassed
Asked to predict exactly what a short program prints. This is the half of coding that matters for agent work and that writing valid syntax does not demonstrate: following state through control flow.
Writes codepassed
Asked for a specific function and checked with the Go compiler’s own parser. Nothing is executed — running model-written code to score a model would be a security hole opened for a metric — but syntactic validity is a fact rather than an opinion.
Follows instructionspassed
Given a plain constraint on its output format and checked for exact obedience. This separates an instruct-tuned model from a base one — a base model answers fluently and ignores every instruction, which then looks like a dozen unrelated weaknesses instead of one.
Sustained outputpassed
Asked to produce a long structured answer and checked for length. A model that stops after a couple of hundred tokens cannot write a report, however capable it is otherwise.
Structured outputpassed
The ENGINE was asked to constrain decoding to a schema, and the answer conformed. This is stronger than politely asking for JSON and hoping: the grammar makes invalid output unreachable, which is what an operator actually builds on.
Extended thinkingpassed
The model emitted reasoning on the engine’s dedicated reasoning channel before answering. Read from that channel rather than by scanning the answer for tags, which is why models that reason are no longer reported as models that do not.
Synthesises sourcespassed
Given three documents with one fact split across them, and required to combine rather than quote. No single document contains the answer, so a model that retrieves without reasoning cannot pass.
Tool callingpassed
The node asked the model to call a function and checked that it called the RIGHT one with a well-formed argument — not merely that it emitted a tool name.
Faithful to tool resultspassed
The model was handed a list through a tool and had to report every item back. Calling a tool and understanding its answer are different abilities: a model can produce a perfect call, receive nine items, and confidently report three. This is the flag that separates them.
Tool loop finishedpassed
Whether the conversation ENDED after the tool call, or the model kept calling until the turn budget ran out. A model that passes tool calling and fails this will call a tool in production and never come back.
Chained tool callspassed
The model had to call one tool, read its answer, and call a second with a value derived from it. One call is not agentic work — real specialist jobs are sequences, and a model that never takes the second step looks, from outside, exactly like a specialist silently doing nothing.
Visionpassed
A four-digit code was rendered into an image and the model had to read it back. A model that cannot see cannot guess four digits, and a model that can has no reason to fail — so this passes and fails for the right reason, which describing a picture never did.
Counts in imagespassed
Shown a row of shapes and asked how many. Reading rendered text and grasping a scene are different abilities, and a model can do the first with no spatial understanding at all — which is what charts, shelf photographs and document layout actually need.
Multiple images at oncepassed
Given two images and asked to compare them. Many engines and models accept one image per turn and silently drop the rest, which turns a comparison into a confident answer about half the evidence.
Spatial reasoningpassed
Asked which of two shapes is larger and on which side — both ways round, so a coin flip cannot pass. Reading glyphs, counting objects and understanding layout are three different abilities, and only the third answers "what is next to what", which is what charts, diagrams and shelf photographs turn on.
Embeddingsprobed, and failed
The engine was asked for an embedding of a short text and returned a real, non-zero vector.
Speech outprobed, and failed
The model itself turned text into audio — identified from the container bytes, not from what the engine labelled them, so an error page announced as audio cannot pass. Distinct from this gram running a separate text-to-speech engine: an operator choosing between them is choosing between two architectures.

Other builds of Gemma 4 E2B

every other build of this member measured on the network — compare what each actually proved
hf.co/unsloth/gemma-4-E2B-it-GGUF:IQ4_XSunsloth · IQ4_XS · ollama
1,31,072 ctx proved2 nodes
gemma4:e2bgoogle · Q4_K_M · ollama
1,31,072 ctx proved8 nodes
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q4_0unsloth · Q4_0 · ollama
1,31,072 ctx proved3 nodes
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q8_0unsloth · Q8_0 · ollama
1,31,072 ctx proved1 node
hf.co/unsloth/gemma-4-E2B-it-GGUF:BF16unsloth · BF16 · ollama
1,31,072 ctx proved1 node
hf.co/unsloth/gemma-4-E2B-it-GGUF:IQ4_NLunsloth · IQ4_NL · ollama
8,192 ctx proved2 nodes
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q3_K_Munsloth · Q3_K_M · ollama
8,192 ctx proved3 nodes
hf.co/unsloth/gemma-4-E2B-it-GGUF:Q3_K_Sunsloth · Q3_K_S · ollama
4,096 ctx proved1 node
gemma4:e2b-it-qatollama · ollama
ctx not probed1 node
google/gemma-4-E2B-it-qat-q4_0-ggufgoogle · llamacpp
ctx not probed1 node
signed by this network

What machines here proved

context proved by recall1,31,072
peak throughput ever signed164.74 tok/s · did:epn:002408…84bf54
resident while loaded4.1 GiB (1.4 on card) · did:epn:002408…53b229
machines that measured it8
samples behind those numbers8
regions it was proved in7
Every line traces to a node signature.
from the public record · unsigned

What the runtime says about it

context declared1,31,072
parameters4.65B
parameter count4,64,74,50,147
quantisationQ4_K_M
familygemma4
formatgguf
ollama /api/show; the node attests that the runtime reported these, not that they are true. Attested by did:epn:002408…a5be46.The only unsigned block on this page. We attest we read these from the runtime — not that they are true. Nobody signs a model card.

Proven across regions

Bhubaneswar, Odisha752115 · 1 node
Kane, Iowa51503 · 1 node
The Dalles, Oregon97058 · 1 node
Bengaluru, Karnataka560037 · 1 node
North 24 Parganas, West Bengal700108 · 2 nodes
Bengaluru, Karnataka562130 · 1 node
Sundargarh, Odisha770033 · 1 node

Every machine that measured it

nothing is averaged — the machine is what you are choosing
did:epn:002408…84bf54Bhubaneswar · 752115 · 0.34.0 · 30 Sept 2026
1,31,072ctx provedprobe QMsgN4+7I7…
87.36tok/s nowpeak 92.56 · 30 Sept 202659 samples · 134s2.7 GiB resident · 2.7 GiB on card · 3.8 GiB on disk · 299.4s cold start
✓ Reasons about code✓ Writes code✓ Follows instructions✓ Sustained output✓ Structured output✓ Extended thinking✓ Synthesises sources✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✗ Counts in images✓ Multiple images at once✓ Vision✓ Spatial reasoning✗ Audio✓ 10 languages
did:epn:002408…e78fd2Kane · 51503 · 0.40.6 · 19 Oct 2026
4,096ctx provedprobe 1O5p3Kj5Ss…
77.58tok/s now35 samples · 98s1.6 GiB resident · 1.6 GiB on card · 3.8 GiB on disk
✓ Reasons about code✓ Writes code✓ Follows instructions✓ Sustained output✓ Structured output✓ Extended thinking✓ Synthesises sources✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✓ Counts in images✓ Multiple images at once✓ Vision✓ Spatial reasoning✗ Audio✓ 10 languages
did:epn:002408…a5be46The Dalles · 97058 · 0.40.10 · on host · 24 Oct 2026Intel Xeon @ 2.00GHz · Tesla T4 15 GiB · 2 cores · 31.3 GiB RAM
1,31,072ctx provedrecalled at every size asked
74.42tok/s now468 samples · 1,464s2.5 GiB resident · 2.5 GiB on card · 3.8 GiB on disk
✗ Speech out✓ Reasons about code✓ Writes code✗ Embeddings✓ Follows instructions✓ Sustained output✓ Structured output✓ Extended thinking✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Counts in images✓ Multiple images at once✓ Vision✓ Spatial reasoning✗ Audio✓ 12 languages
did:epn:002408…53b229Bengaluru · 560037 · ollama-host · 12 Oct 2026
0ctx provedno recall at 8,192 · 4,096
27.39tok/s nowpeak 38.87 · 17 Sept 2026247 samples · 1,962s4.1 GiB resident · 1.4 GiB on card · 3.8 GiB on disk · 25.0s cold start
✓ Structured output✓ Extended thinking✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✓ Vision✗ Audio
did:epn:002408…9a1db0North 24 Parganas · 700108 · ollama · 16 Sept 2026
0ctx provedno recall at 1,31,072 · 65,536 · 32,768 · 16,384 · 8,192 · 4,096
15.33tok/s now10 samples · 116s
✓ Reasons about code✓ Writes code✓ Follows instructions✓ Sustained output✓ Structured output✓ Extended thinking✓ Synthesises sources✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✗ Vision✗ Audio✓ 10 languages
did:epn:002408…12954dNorth 24 Parganas · 700108 · ollama · 16 Sept 2026
0ctx provedno recall at 1,31,072 · 65,536 · 32,768 · 16,384 · 8,192 · 4,096
9.95tok/s nowpeak 10.64 · 16 Sept 202616 samples · 213s3.8 GiB on disk
✓ Reasons about code✓ Writes code✓ Follows instructions✓ Sustained output✓ Structured output✓ Extended thinking✓ Synthesises sources✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✓ Counts in images✓ Multiple images at once✓ Vision✓ Spatial reasoning✗ Audio✓ 10 languages
did:epn:002408…9e3c79Bengaluru · 562130 · ollama · 14 Sept 2026Intel64 Family 6 Model 151 Stepping 5, GenuineIntel · 15.7 GiB RAM
0ctx provedno recall at 4,096
6.80tok/s now47 samples · 955s
✓ Structured output✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✗ Extended thinking✗ Vision✗ Audio
did:epn:002408…d68672Sundargarh · 770033 · ollama-host · on host · 22 Oct 2026Intel Core i5-5350U @ 1.80GHz · 2 cores · 8.0 GiB RAM
0ctx provedthis machine could not serve at 4K
3.99tok/s now70 samples · 3,033s
✓ Structured output✓ Tool calling✓ Faithful to tool results✓ Tool loop finished✓ Chained tool calls✗ Extended thinking✗ Vision✗ Audio

Context ladders attempted: did:epn:002408…84bf54 131072,65536,32768,16384,8192,4096 · did:epn:002408…e78fd2 4096 · did:epn:002408…a5be46 4096,8192,16384,32768,65536,131072 · did:epn:002408…53b229 8192,4096 · did:epn:002408…9a1db0 131072,65536,32768,16384,8192,4096 · did:epn:002408…12954d 131072,65536,32768,16384,8192,4096 · did:epn:002408…9e3c79 4096 · did:epn:002408…d68672 4096

Run this where your data already is

Tell us the workload, whose machines it should run on, and which jurisdiction it answers to.

Run a gramPut a machine on the network in this region; it signs its capacity into the room.Open a gramxStart your own room — an org, a community, a brand.Be invitedAccept a link from an admin to cross into the private room.