[GH-ISSUE #1749] The "seed" is not working reliable for me. #1000

New Issue

GiteaMirror · 2026-04-12T10:42:39-05:00

GiteaMirror commented

2026-04-12 10:42:39 -05:00

Originally created by @oderwat on GitHub (Dec 30, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1749

Originally assigned to: @BruceMacD on GitHub.

I am using a seed (int 1) for prompt generation with a mistral model, and it works not reliable. Instead, I get some interesting results with a pattern:

EDIT: It seems like this behavior is independent of the seed choice and the seeds are not working at all?

When freshly start ollama serve and send the exact same prompt together with a seed to "/api/generate" (stream: false) I always get three times the same reply. The fourth and all following replies are then different!

When I switch the model and make the same prompt to that, it also gives three of the same and then varying results!

As a workaround, I actually switch to another very small model and create a minimal embedding (is faster than doing an inference prompt) before doing the actual prompt and this gives me reliable results. Even if that actually works and is quite fast in my case, I think there is a problem that needs to be fixed.

I am on the current main 2a2fa3c

Originally created by @oderwat on GitHub (Dec 30, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1749 Originally assigned to: @BruceMacD on GitHub. I am using a seed (int 1) for prompt generation with a mistral model, and it works not reliable. Instead, I get some interesting results with a pattern: EDIT: It seems like this behavior is independent of the seed choice and the seeds are not working at all? When freshly start `ollama serve` and send the exact same prompt together with a seed to "/api/generate" (stream: false) I always get three times the same reply. The fourth and all following replies are then different! When I switch the model and make the same prompt to that, it also gives three of the same and then varying results! As a workaround, I actually switch to another very small model and create a minimal embedding (is faster than doing an inference prompt) before doing the actual prompt and this gives me reliable results. Even if that actually works and is quite fast in my case, I think there is a problem that needs to be fixed. I am on the current main [2a2fa3c](https://github.com/jmorganca/ollama/commit/2a2fa3c3298194f4f3790aade78df2f53d170d8e)

GiteaMirror added the bug label 2026-04-12 10:42:39 -05:00

GiteaMirror closed this issue

2026-04-12 10:42:40 -05:00

GiteaMirror commented

2026-04-12 10:42:41 -05:00

@technovangelist commented on GitHub (Jan 2, 2024):

Interesting, I tried out this code:

async function test() {
	const body = {
		"model": "mistral", 
		"prompt": "list 3 synonyms for a sink", 
		"stream": false, 
		"options": {
			"seed": 12345, 
			"temperature": 0
		}
	};
	const response = await fetch("http://localhost:11434/api/generate", {
		"method": "POST", 
		"body": JSON.stringify(body), 
	});
	
	const out = await response.json();
	
	console.log(out.response)
}

test();
test();
test();
test();
test();
test();
test();
test();
test();
test();

When I leave out the temperature, I get somewhat random responses each time. But use the seed and temp and I get the same results every single time.

@technovangelist commented on GitHub (Jan 2, 2024): Interesting, I tried out this code: ``` async function test() { const body = { "model": "mistral", "prompt": "list 3 synonyms for a sink", "stream": false, "options": { "seed": 12345, "temperature": 0 } }; const response = await fetch("http://localhost:11434/api/generate", { "method": "POST", "body": JSON.stringify(body), }); const out = await response.json(); console.log(out.response) } test(); test(); test(); test(); test(); test(); test(); test(); test(); test(); ``` When I leave out the temperature, I get somewhat random responses each time. But use the seed and temp and I get the same results every single time.

GiteaMirror commented

2026-04-12 10:42:42 -05:00

@BruceMacD commented on GitHub (Jan 2, 2024):

The suggestion above (using temperature) helps a lot, although I do occasionally see some variation here after may requests.

Some other things to note:

There was a bug in main where seed wouldn't be set properly, which is fixed in #1761. If you weren't building from source this will not have been the issue.
If you're setting temperate in the interactive ollama run the chat history will effect what is generated also.

@BruceMacD commented on GitHub (Jan 2, 2024): The suggestion above (using temperature) helps a lot, although I do occasionally see some variation here after may requests. Some other things to note: - There was a bug in `main` where seed wouldn't be set properly, which is fixed in #1761. If you weren't building from source this will not have been the issue. - If you're setting `temperate` in the interactive `ollama run` the chat history will effect what is generated also.

GiteaMirror commented

2026-04-12 10:42:42 -05:00

@oderwat commented on GitHub (Jan 2, 2024):

I recompiled from main with the mentioned PR merged and basically have the same result as before (I thought it was different, but I still had my "reset" workaround active).

@technovangelist have you tried other seeds?

@oderwat commented on GitHub (Jan 2, 2024): I recompiled from main with the mentioned PR merged and basically have the same result as before (I thought it was different, but I still had my "reset" workaround active). @technovangelist have you tried other seeds?

GiteaMirror commented

2026-04-12 10:42:42 -05:00

@oderwat commented on GitHub (Jan 2, 2024):

@technovangelist This is what I get using the current main and your little script (pointing at my server and using 'dolphin2.2-mistral:7b-q4_K_M')

This is after restarting the ollama server and using a different seed:

P.S.: The ollama server runs on WSL 2 using a RTX 3090 and 64 GB RAM.

EDIT: I started the ollama server on my iMac with the same model and script. There it replies twice with the identical and on the third call starts to deviate. Also, independent of the seed I use.

@oderwat commented on GitHub (Jan 2, 2024): @technovangelist This is what I get using the current main and your little script (pointing at my server and using 'dolphin2.2-mistral:7b-q4_K_M') <img width="909" alt="image" src="https://github.com/jmorganca/ollama/assets/719156/13efced0-73eb-4fb5-9c21-aca30bd3d79d"> This is after restarting the ollama server and using a different seed: <img width="234" alt="image" src="https://github.com/jmorganca/ollama/assets/719156/2ce2659f-d707-44c2-bcba-2334dd86e5a7"> P.S.: The ollama server runs on WSL 2 using a RTX 3090 and 64 GB RAM. EDIT: I started the ollama server on my iMac with the same model and script. There it replies twice with the identical and on the third call starts to deviate. Also, independent of the seed I use.

GiteaMirror commented

2026-04-12 10:42:43 -05:00

@oderwat commented on GitHub (Jan 3, 2024):

Using the newest main (a00367b2f92) I still can't get the seed to work reliable. But I wonder if that is "just me" or if this is a confirmed problem. From all my experiments it looks like as if there is no real support for a 'seed' or implemented in an unusable way.

@oderwat commented on GitHub (Jan 3, 2024): Using the newest main (a00367b2f92) I still can't get the seed to work reliable. But I wonder if that is "just me" or if this is a confirmed problem. From all my experiments it looks like as if there is no real support for a 'seed' or implemented in an unusable way.

GiteaMirror commented

2026-04-12 10:42:43 -05:00

@technovangelist commented on GitHub (Jan 4, 2024):

Thanks @oderwat for checking. I know @BruceMacD is looking into it. We added a bug label to the issue so we will continue investigating.

@technovangelist commented on GitHub (Jan 4, 2024): Thanks @oderwat for checking. I know @BruceMacD is looking into it. We added a bug label to the issue so we will continue investigating.

GiteaMirror commented

2026-04-12 10:42:44 -05:00

@jmorganca commented on GitHub (Feb 20, 2024):

@oderwat for consistent output both seed must be set to a given number, and temperature must be set to 0. Let me know if this doesn't help.

I've added some examples of reproducible outputs to api.md:

https://github.com/ollama/ollama/blob/main/docs/api.md#request-reproducible-outputs

@jmorganca commented on GitHub (Feb 20, 2024): @oderwat for consistent output both `seed` must be set to a given number, and `temperature` must be set to 0. Let me know if this doesn't help. I've added some examples of reproducible outputs to `api.md`: https://github.com/ollama/ollama/blob/main/docs/api.md#request-reproducible-outputs

GiteaMirror commented

2026-04-12 10:42:44 -05:00

@d-kleine commented on GitHub (Jun 27, 2024):

@jmorganca To make the output on the same OS reproducible, you need to set a fixed context length with num_ctx too (otherwise the output slightly differs when not setting it)

Also, the having set seed, temperature to 0 and num_ctx, the output is deterministic and reproducible, but inconsistent over different OS. Might be related to llama.cpp: https://github.com/ggerganov/llama.cpp/discussions/2100

@d-kleine commented on GitHub (Jun 27, 2024): @jmorganca To make the output on the same OS reproducible, you need to set a fixed context length with `num_ctx` too (otherwise the output slightly differs when not setting it) Also, the having set `seed`, `temperature` to 0 and `num_ctx`, the output is deterministic and reproducible, but inconsistent over different OS. Might be related to llama.cpp: https://github.com/ggerganov/llama.cpp/discussions/2100

GiteaMirror commented

2026-04-12 10:42:45 -05:00

@jtyska commented on GitHub (Nov 13, 2024):

Hey! I'm still having this problem. When I set temperature to 0 and use different seed values, the generated content is always the same. Any clue on how to fix it? Here is the payload I'm using to generate content via API (/generate)

"model": "qwen2.5:72b",
"prompt": "Generate something",
"options":{
"seed": 99, # here I change to 100, 101, 102, 103 anyvalue and the generated content is always the same
"temperature":0
},
"stream": False

        My ollama version is 0.4.1 got from the docker image on dockerhub.
        
       Could anyone give me a hand on this?

@jtyska commented on GitHub (Nov 13, 2024): Hey! I'm still having this problem. When I set temperature to 0 and use different seed values, the generated content is always the same. Any clue on how to fix it? Here is the payload I'm using to generate content via API (/generate) "model": "qwen2.5:72b", "prompt": "Generate something", "options":{ "seed": 99, # here I change to 100, 101, 102, 103 anyvalue and the generated content is always the same "temperature":0 }, "stream": False My ollama version is 0.4.1 got from the docker image on dockerhub. Could anyone give me a hand on this?

GiteaMirror commented

2026-04-12 10:42:45 -05:00

@d-kleine commented on GitHub (Nov 13, 2024):

@jtyska You probably need to pass a fixed context length too as this is not set up in the template: https://ollama.com/library/qwen2:72b/blobs/f02dd72bb242. The reason is that the context length will be random then too, leading the different lengths (and therefore also tokens) generated.

So try this by adding:

...
"num_ctx": 32_768
...

to your code and let us know is this fixes the issue for you.

@d-kleine commented on GitHub (Nov 13, 2024): @jtyska You probably need to pass a fixed context length too as this is not set up in the template: https://ollama.com/library/qwen2:72b/blobs/f02dd72bb242. The reason is that the context length will be random then too, leading the different lengths (and therefore also tokens) generated. So try this by adding: ```python ... "num_ctx": 32_768 ... ``` to your code and let us know is this fixes the issue for you.

GiteaMirror commented

2026-04-12 10:42:45 -05:00

@oderwat commented on GitHub (Nov 13, 2024):

It sounds counterintuitive that making more stuff "fixed" will suddenly create different outputs when they were the same with different seeds before. Could you elaborate on why that is?

The reason is that the context length will be random then too, leading the different lengths (and therefore also tokens) generated.

?

And, why would the context be random anyway? The context of a model is a fixed value. What may change is what you feed to the model by changing shortening the prompt. What am I missing?

@oderwat commented on GitHub (Nov 13, 2024): It sounds counterintuitive that making more stuff "fixed" will suddenly create different outputs when they were the same with different seeds before. Could you elaborate on why that is? > The reason is that the context length will be random then too, leading the different lengths (and therefore also tokens) generated. ? And, why would the context be random anyway? The context of a model is a fixed value. What may change is what you feed to the model by changing shortening the prompt. What am I missing?

GiteaMirror commented

2026-04-12 10:42:45 -05:00

@d-kleine commented on GitHub (Nov 13, 2024):

@oderwat Sorry, you are right – I was juggling too many things at once and got confused. The context length is indeed fixed, and what changes is the input or output length based on the prompt or generation settings.

@jtyska I just tested the settings in Ollama, you should be able to get consistent outputs per seed with the seed settings. If you you want variation across different seeds, you need to increase the temperature to a value >0 (for instance 0.7). So your settings would be:

{
        "model": "qwen2.5:70b",
        "messages": [
            {
                "role": "user",
                "content":  "Generate something"
            }
        ],
        "options": {
            "seed": 101,
            "temperature": 0.7            
        },
        "stream": False
    }

Then every generated output is consistent when using the same seed, but varies across different seeds.

@d-kleine commented on GitHub (Nov 13, 2024): @oderwat Sorry, you are right – I was juggling too many things at once and got confused. The context length is indeed fixed, and what changes is the input or output length based on the prompt or generation settings. @jtyska I just tested the settings in Ollama, you should be able to get consistent outputs per seed with the `seed` settings. If you you want variation across different seeds, you need to increase the temperature to a value >0 (for instance `0.7`). So your settings would be: ```python { "model": "qwen2.5:70b", "messages": [ { "role": "user", "content": "Generate something" } ], "options": { "seed": 101, "temperature": 0.7 }, "stream": False } ``` Then every generated output is consistent when using the same seed, but varies across different seeds.

GiteaMirror commented

2026-04-12 10:42:46 -05:00

@jtyska commented on GitHub (Nov 13, 2024):

If you increase the temperature, reproducibility will be impossible. The answer for the same seed, with temperature 0, should be the same, but different seeds, with temperature zero, should produce different responses. Isn't that right? Otherwise, what's the point of having a seed parameter?

@jtyska commented on GitHub (Nov 13, 2024): If you increase the temperature, reproducibility will be impossible. The answer for the same seed, with temperature 0, should be the same, but different seeds, with temperature zero, should produce different responses. Isn't that right? Otherwise, what's the point of having a seed parameter?

GiteaMirror commented

2026-04-12 10:42:46 -05:00

@d-kleine commented on GitHub (Nov 13, 2024):

Please try out the settings I have provided, and let me know if that worked for you.

If you increase the temperature, reproducibility will be impossible. The answer for the same seed, with temperature 0, should be the same, but different seeds, with temperature zero, should produce different responses. Isn't that right? Otherwise, what's the point of having a seed parameter?

Regarding your question, that's not right - the seed initializes a random number generator used during text generation, changing which tokens will be chosen (i.e., for temperature > 0). A fixed seed ensures consistency and reproducibility in such cases. However, when temperature = 0, the model becomes fully deterministic (=always selecting the most likely tokens), meaning there is no randomness in token selection and changing the seed will not affect the output then.

@d-kleine commented on GitHub (Nov 13, 2024): Please try out the settings I have provided, and let me know if that worked for you. > If you increase the temperature, reproducibility will be impossible. The answer for the same seed, with temperature 0, should be the same, but different seeds, with temperature zero, should produce different responses. Isn't that right? Otherwise, what's the point of having a seed parameter? Regarding your question, that's not right - the seed initializes a random number generator used during text generation, changing which tokens will be chosen (i.e., for **temperature > 0**). A fixed seed ensures consistency and reproducibility in such cases. However, when **temperature = 0**, the model becomes fully deterministic (=always selecting the most likely tokens), meaning there is no randomness in token selection and changing the seed will not affect the output then.

GiteaMirror commented

2026-04-12 10:42:46 -05:00

@oderwat commented on GitHub (Nov 13, 2024):

@d-kleine That would mean that with a fixed seed the result of a fixed temperature should be deterministic?

@oderwat commented on GitHub (Nov 13, 2024): @d-kleine That would mean that with a fixed seed the result of a fixed temperature should be deterministic?

GiteaMirror commented

2026-04-12 10:42:47 -05:00

@d-kleine commented on GitHub (Nov 13, 2024):

Depends on what you mean by "fixed temperature". So for temp = 0 (in practice, it's a very small number close to zero), yes; And for temp > 0, only under ideal conditions (there can be different calculations due to underlying hardware differences or model architecture complexities though)

I think it's easier to distinguish using consistency and determinism:

fixed seed: ensures consistency when randomness is involved
temp=0: guarantees determinism

So, both settings contribute to reproducibility but from different sides: the seed controls randomness when it's present, and temperature controls whether randomness exists at all.

@d-kleine commented on GitHub (Nov 13, 2024): Depends on what you mean by "fixed temperature". So for temp = 0 (in practice, it's a very small number close to zero), yes; And for temp > 0, only under ideal conditions (there can be different calculations due to underlying hardware differences or model architecture complexities though) I think it's easier to distinguish using **consistency** and **determinism**: * fixed seed: ensures **consistency** when randomness is involved * temp=0: guarantees **determinism** So, both settings contribute to reproducibility but from different sides: the seed controls randomness when it's present, and temperature controls whether randomness exists at all.

GiteaMirror commented

2026-04-12 10:42:47 -05:00

@oderwat commented on GitHub (Nov 13, 2024):

"temp=0: guarantees determinism"

Aha? Between different models and hardware? Really?

And of course we all here talk about the same model, same hardware (and a temperature > 0.0) with the same seed vs. another seed.

What I think what happens is the following:

The random generator used is not a (deterministic) pseudo random generator and that means the seed is useless if s more than one number needed.

Could that be the case?

@oderwat commented on GitHub (Nov 13, 2024): > "temp=0: guarantees determinism" Aha? Between different models and hardware? Really? And of course we all here talk about the same model, same hardware (and a temperature > 0.0) with the same seed vs. another seed. What I think what happens is the following: The random generator used is not a (deterministic) pseudo random generator and that means the seed is useless if s more than one number needed. Could that be the case?

GiteaMirror commented

2026-04-12 10:42:47 -05:00

@d-kleine commented on GitHub (Nov 13, 2024):

Aha? Between different models and hardware? Really?

Yes, I was actually involved in some of those discussions around it, see for example https://github.com/ollama/ollama/pull/5760#issuecomment-2462917232 and https://github.com/ggerganov/llama.cpp/issues/8353#issuecomment-2304407109

What I think what happens is the following:

The random generator used is not a (deterministic) pseudo random generator and that means the seed is useless if s more than one number needed.

Could that be the case?

In this case, no - most modern LLMs (like Qwen) use deterministic PRNG.

@d-kleine commented on GitHub (Nov 13, 2024): > Aha? Between different models and hardware? Really? Yes, I was actually involved in some of those discussions around it, see for example https://github.com/ollama/ollama/pull/5760#issuecomment-2462917232 and https://github.com/ggerganov/llama.cpp/issues/8353#issuecomment-2304407109 > What I think what happens is the following: > > The random generator used is not a (deterministic) pseudo random generator and that means the seed is useless if s more than one number needed. > > Could that be the case? In this case, no - most modern LLMs (like Qwen) use deterministic PRNG.

GiteaMirror commented

2026-04-12 10:42:47 -05:00

@jtyska commented on GitHub (Nov 13, 2024):

Thanks a lot @d-kleine I finally got it. I tested with a higher temperature and the same seed, and it indeed produced the same output!

@jtyska commented on GitHub (Nov 13, 2024): Thanks a lot @d-kleine I finally got it. I tested with a higher temperature and the same seed, and it indeed produced the same output!

GiteaMirror commented

2026-04-12 10:42:48 -05:00

@filijamz commented on GitHub (Mar 1, 2025):

I tried some experiments with llama3.2 with no memory (history) and random seed values. I kept asking for a riddle expecting a new riddle with a different seed each time. Instead, it responded with the exact same riddle each time. Am I misunderstanding what seed does?

@filijamz commented on GitHub (Mar 1, 2025): I tried some experiments with llama3.2 with no memory (history) and random seed values. I kept asking for a riddle expecting a new riddle with a different seed each time. Instead, it responded with the exact same riddle each time. Am I misunderstanding what seed does?

GiteaMirror referenced this issue

2026-04-12 23:00:26 -05:00

[PR #1000] [CLOSED] Added clear command #10430

GiteaMirror referenced this issue

2026-04-16 05:05:13 -05:00

[PR #1000] [CLOSED] Added clear command #15701

GiteaMirror referenced this issue

2026-04-19 15:21:48 -05:00

[PR #1000] [CLOSED] Added clear command #20970

GiteaMirror referenced this issue

2026-04-22 20:59:03 -05:00

[PR #1000] [CLOSED] Added clear command #36303

GiteaMirror referenced this issue

2026-04-24 21:31:40 -05:00

[PR #1000] [CLOSED] Added clear command #41678

GiteaMirror referenced this issue

2026-04-29 11:42:12 -05:00

[PR #1000] [CLOSED] Added clear command #57127

GiteaMirror referenced this issue

2026-05-05 04:16:19 -05:00

[PR #1000] [CLOSED] Added clear command #72724

Sign in to join this conversation.

Branches Tags

main

dhiltgen/ci

parth-launch-plan-gating

hoyyeva/anthropic-reference-images-path

parth-anthropic-reference-images-path

brucemacd/download-before-remove

hoyyeva/editor-config-repair

parth-mlx-decode-checkpoints

parth-launch-codex-app

hoyyeva/fix-codex-model-metadata-warning

hoyyeva/qwen

parth/hide-claude-desktop-till-release

hoyyeva/opencode-image-modality

parth-add-claude-code-autoinstall

release_v0.22.0

pdevine/manifest-list

codex/fix-codex-model-metadata-warning

pdevine/addressable-manifest

brucemacd/launch-fetch-reccomended

jmorganca/llama-compat

launch-copilot-cli

hoyyeva/opencode-thinking

release_v0.20.7

parth-auto-save-backup

parth-test

jmorganca/gemma4-audio-replacements

fix-manifest-digest-on-pull

hoyyeva/vscode-improve

brucemacd/install-server-wait

parth/update-claude-docs

brucemac/start-ap-install

pdevine/mlx-update

pdevine/qwen35_vision

drifkin/api-show-fallback

mintlify/image-generation-1773352582

hoyyeva/server-context-length-local-config

jmorganca/faster-reptition-penalties

jmorganca/convert-nemotron

parth-pi-thinking

pdevine/sampling-penalties

jmorganca/fix-create-quantization-memory

dongchen/resumable_transfer_fix

pdevine/sampling-cache-error

jessegross/mlx-usage

hoyyeva/openclaw-config

hoyyeva/app-html

pdevine/qwen3next

brucemacd/sign-sh-install

brucemacd/tui-update

brucemacd/usage-api

jmorganca/launch-empty

fix-app-dist-embed

mxyng/mlx-compile

mxyng/mlx-quant

mxyng/mlx-glm4.7

mxyng/mlx

brucemacd/simplify-model-picker

jmorganca/qwen3-concurrent

fix-glm-4.7-flash-mla-config

drifkin/qwen3-coder-opening-tag

brucemacd/usage-cli

fix-cuda12-fattn-shmem

ollama-imagegen-docs

parth/fix-multiline-inputs

brucemacd/config-docs

mxyng/model-files

mxyng/simple-execute

fix-imagegen-ollama-models

mxyng/async-upload

jmorganca/lazy-no-dtype-changes

imagegen-auto-detect-create

parth/decrease-concurrent-download-hf

fix-mlx-quantize-init

jmorganca/x-cleanup

usage

imagegen-readme

jmorganca/glm-image

mlx-gpu-cd

jmorganca/imagegen-modelfile

parth/agent-skills

parth/agent-allowlist

parth/signed-in-offline

parth/agents

parth/fix-context-chopping

improve-cloud-flow

parth/add-models-websearch

parth/prompt-renderer-mcp

jmorganca/native-settings

jmorganca/download-stream-hash

jmorganca/client2-rebased

brucemacd/oai-chat-req-multipart

jessegross/multi_chunk_reserve

grace/additional-omit-empty

grace/mistral-3-large

mxyng/tokenizer2

mxyng/tokenizer

jessegross/flash

hoyyeva/windows-nacked-app

mxyng/cleanup-attention

grace/deepseek-parser

hoyyeva/remember-unsent-prompt

parth/add-lfs-pointer-error-conversion

parth/olmo2-test2

hoyyeva/ollama-launchagent-plist

nicole/olmo-model

parth/olmo-test

mxyng/remove-embedded

parth/render-template

jmorganca/intellect-3

parth/remove-prealloc-linter

jmorganca/cmd-eval

nicole/nomic-embed-text-fix

mxyng/lint-2

hoyyeva/add-gemini-3-pro-preview

hoyyeva/load-model-list

mxyng/expand-path

mxyng/environ-2

hoyyeva/deeplink-json-encoding

parth/improve-tool-calling-tests

hoyyeva/conversation

hoyyeva/assistant-edit-response

hoyyeva/thinking

origin/brucemacd/invalid-char-i-err

parth/improve-tool-calling

jmorganca/required-omitempty

grace/qwen3-vl-tests

mxyng/iter-client

parth/docs-readme

nicole/embed-test

pdevine/integration-benchstat

parth/remove-generate-cmd

parth/add-toolcall-id

mxyng/server-tests

jmorganca/glm-4.6

jmorganca/gin-h-compat

drifkin/stable-tool-args

pdevine/qwen3-more-thinking

parth/add-websearch-client

nicole/websearch_local

jmorganca/qwen3-coder-updates

grace/deepseek-v3-migration-tests

mxyng/fix-create

jmorganca/cloud-errors

pdevine/parser-tidy

revert-12233-parth/simplify-entrypoints-runner

parth/enable-so-gpt-oss

brucemacd/qwen3vl

jmorganca/readme-simplify

parth/gpt-oss-structured-outputs

revert-12039-jmorganca/tools-braces

mxyng/embeddings

mxyng/gguf

mxyng/benchmark

mxyng/types-null

parth/move-parsing

mxyng/gemma2

jmorganca/docs

mxyng/16-bit

mxyng/create-stdin

pdevine/authorizedkeys

mxyng/quant

parth/opt-in-error-context-window

brucemacd/cache-models

brucemacd/runner-completion

jmorganca/llama-update-6

brucemacd/benchmark-list

brucemacd/partial-read-caps

parth/deepseek-r1-tools

mxyng/omit-array

parth/tool-prefix-temp

brucemacd/runner-test

jmorganca/qwen25vl

brucemacd/model-forward-test-ext

parth/python-function-parsing

jmorganca/cuda-compression-none

drifkin/num-parallel

drifkin/chat-truncation-fix

jmorganca/sync

parth/python-tools-calling

drifkin/array-head-count

brucemacd/create-no-loop

parth/server-enable-content-stream-with-tools

qwen25omni

mxyng/v3

brucemacd/ropeconfig

jmorganca/silence-tokenizer

parth/sample-so-test

parth/sampling-structured-outputs

brucemacd/doc-go-engine

parth/constrained-sampling-json

jmorganca/mistral-wip

brucemacd/mistral-small-convert

parth/sample-unmarshal-json-for-params

brucemacd/jomorganca/mistral

pdevine/bfloat16

jmorganca/mistral

brucemacd/mistral

pdevine/logging

parth/sample-correctness-fix

parth/sample-fix-sorting

jmorgan/sample-fix-sorting-extras

jmorganca/temp-0-images

brucemacd/parallel-embed-models

brucemacd/shim-grammar

jmorganca/fix-gguf-error

bmizerany/nameswork

jmorganca/faster-releases

bmizerany/validatenames

brucemacd/err-no-vocab

brucemacd/rope-config

brucemacd/err-hint

brucemacd/qwen2_5

brucemacd/logprobs

brucemacd/new_runner_graph_bench

progress-flicker

brucemacd/forward-test

brucemacd/go_qwen2

pdevine/gemma2

jmorganca/add-missing-symlink-eval

mxyng/next-debug

parth/set-context-size-openai

brucemacd/next-bpe-bench

brucemacd/next-bpe-test

brucemacd/new_runner_e2e

brucemacd/new_runner_qwen2

pdevine/convert-cohere2

brucemacd/convert-cli

parth/log-probs

mxyng/next-mlx

mxyng/cmd-history

parth/templating

parth/tokenize-detokenize

brucemacd/check-key-register

bmizerany/grammar

jmorganca/vendor-081b29bd

mxyng/func-checks

jmorganca/fix-null-format

parth/fix-default-to-warn-json

jmorganca/qwen2vl

jmorganca/no-concat

parth/cmd-cleanup-SO

brucemacd/check-key-register-structured-err

parth/openai-stream-usage

parth/fix-referencing-so

stream-tools-stop

jmorganca/degin-1

brucemacd/install-path-clean

brucemacd/push-name-validation

brucemacd/browser-key-register

jmorganca/openai-fix-first-message

jmorganca/fix-proxy

jessegross/sample

parth/disallow-streaming-tools

dhiltgen/remove_submodule

jmorganca/ga

jmorganca/mllama

pdevine/newlines

pdevine/geems-2b

jmorganca/llama-bump

mxyng/modelname-7

mxyng/gin-slog

mxyng/modelname-6

jyan/convert-prog

jyan/quant5

paligemma-support

pdevine/import-docs

jmorganca/openai-context

jyan/paligemma

jyan/p2

jyan/palitest

bmizerany/embedspeedup

jmorganca/llama-vit

brucemacd/allow-ollama

royh/ep-methods

royh/whisper

mxyng/api-models

mxyng/fix-memory

jyan/q4_4/8

jyan/ollama-v

royh/stream-tools

roy-embed-parallel

bmizerany/hrm

revert-5963-revert-5924-mxyng/llama3.1-rope

royh/embed-viz

jyan/local2

jyan/auth

jyan/local

jyan/parse-temp

jmorganca/template-mistral

jyan/reord-g

royh-openai-suffixdocs

royh-imgembed

royh-embed-parallel

jyan/quant4

royh-precision

jyan/progress

pdevine/fix-template

jyan/quant3

pdevine/ggla

mxyng/update-registry-domain

jmorganca/ggml-static

mxyng/create-context

jyan/v0.146

mxyng/layers-from-files

build_dist

bmizerany/noseek

royh-ls

royh-name

timeout

mxyng/server-timestamp

bmizerany/nosillyggufslurps

royh-params

jmorganca/llama-cpp-7c26775

royh-openai-delete

royh-show-rigid

jmorganca/enable-fa

jmorganca/no-error-template

jyan/format

royh-testdelete

bmizerany/fastverify

language_support

pdevine/ps-glitches

brucemacd/tokenize

bruce/iq-quants

bmizerany/filepathwithcoloninhost

mxyng/split-bin

bmizerany/client-registry

jmorganca/if-none-match

native

jmorganca/native

jmorganca/batch-embeddings

jmorganca/initcmake

jmorganca/mm

pdevine/showggmlinfo

modenameenforcealphanum

bmizerany/modenameenforcealphanum

jmorganca/done-reason

jmorganca/llama-cpp-8960fe8

ollama.com

bmizerany/filepathnobuild

bmizerany/types/model/defaultfix

rmdisplaylong

nogogen

bmizerany/x

modelfile-readme

bmizerany/replacecolon

jmorganca/limit

jmorganca/execstack

jmorganca/replace-assets

mxyng/tune-concurrency

jmorganca/testing

whitespace-detection

jmorganca/options

upgrade-all

scratch

cuda-search

mattw/airenamer

mattw/allmodelsonhuggingface

mattw/quantcontext

mattw/whatneedstorun

brucemacd/llama-mem-calc

mattw/faq-context

mattw/communitylinks

mattw/noprune

mattw/python-functioncalling

rename

mxyng/install

pulse

remove-first

editor

mattw/selfqueryingretrieval

cgo

mattw/howtoquant

api

matt/streamingapi

format-config

mxyng/extra-args

shell

update-nous-hermes

cp-model

upload-progress

fix-unknown-model

fix-model-names

delete-fix

insecure-registry

ls

deletemodels

progressbar

readme-updates

license-layers

skip-list

list-models

modelpath

matt/examplemodelfiles

distribution

go-opts

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: github-starred/ollama#1000