Model Follower.

Up to date

Rebuilt
2026-10-25 06:54 UTC
Release radar
117 on watch
Hub snapshot
2026-10-25
Re-checked
daily · every number sourced and dated

Models › Ornith-1.5 family

Ornith-1.5-9B

Hugging Face ↗ Follow changes

ornith-ai MIT out 2 months ago

9.65B parameters, text in, text out, 256K tokens of context. 1 provider serves it.

Verified 2026-10-25 Rank #602 in LLMs · score 74 method v1 Sources 3 Downloads 30d 460.7k · #171 of 619 tracked

Changed this week: 10 new quantized builds on the Hub (97 in all).2026-10-24 · 9 new quantized builds on the Hub (96 in all).2026-10-24

Can I run it?5.8GB at INT4Estimatefits a 16 GB cardmethod
Can I use it?MITPermissivelicence file2026-10-24
Do people use it?460.7kdownloads, 30 days#171 of 619 we track · top 28%+11 likes, downloads +4% this week392 likesHugging Face2026-10-24

Worth reading

Picked by rule from 1 tracked piece: the lab's own post, the most shared, a field report, the latest.

Latest · 2 months ago

Ornith-1.5: From Self-Scaffolding to Self-Improvement

Ornith-1.5 extends its self-scaffolding framework into a complete self-improvement loop that jointly optimizes task generation, scaffold construction, and solution rollouts.

  • The 9B model can be deployed on iPhone and Android devices while outperforming larger models like Gemma 4-31B.
  • The system proposes progressively harder tasks that go beyond what the model has already solved to expose capability gaps.
ornith.ai 2026-08-19 216 HN pts · 76 comments 2.3k words about the Ornith 1.5 family family

Where to run it

Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.

#Cheapest live rowIn /1MOut /1MContextQuantSeen
1 Featherless — — — — Oct 25

All 1 provider →

Where it's used

Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.

Served by 1 API provider; heavily run locally (97 quantized builds); a big base for builders (62 fine-tunes, adapters and merges).

  • As a service1Featherless
  • On own hardware97quantized builds
  • Built on6240 fine-tunes · 9 adapters · 13 merges
  • In apps4Hugging Face Spaces

Will it fit?

Weights only, an estimate: leave room for the KV cache. Method

Fits a 16 GB card at INT4

About 5.8 GB at INT4, 12 GB at FP8, 23 GB at BF16: weights only, so leave room for the KV cache.

16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 5.8 GB of weights (estimate; KV cache and activations extra)INT45.8 GBFP8: about 12 GB of weights (estimate; KV cache and activations extra)FP812 GBBF16: about 23 GB of weights (estimate; KV cache and activations extra)BF1623 GB
16 GB card64 GB Mac2×80 GB8×80 GBINT4: about 5.8 GB of weights (estimate; KV cache and activations extra)INT45.8 GBFP8: about 12 GB of weights (estimate; KV cache and activations extra)FP812 GBBF16: about 23 GB of weights (estimate; KV cache and activations extra)BF1623 GB

Estimate params × bytes per param × 1.2 · how we estimate

The story so far

Weights, every dated article, and repo commits on one rail.

1 dated article and 29 repo commits so far

Teal is the weights (2026-08-18), amber triangles are 1 dated article (each links to it), grey ticks are repo commits.

SepOct2026-08-23: 3 repo commits2026-08-20: 1 repo commit2026-08-19: 22 repo commits2026-08-18: 3 repo commitsOrnith-1.5: From Self-Scaffolding to Self-Improvement · 2026-08-19Weights 2026-08-18 · 2026-08-18Weights 2026-08-18now · 1 provider
SepOct2026-08-23: 3 repo commits2026-08-20: 1 repo commit2026-08-19: 22 repo commits2026-08-18: 3 repo commitsOrnith-1.5: From Self-Scaffolding to Self-Improvement · 2026-08-19Weights 2026-08-18 · 2026-08-18Weights 2026-08-18now · 1 provider

What people built on it

159 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.

159 derivatives on the Hub, most of them quantized

From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.

97 quantized (Hugging Face model tree)40 fine-tunes (Hugging Face model tree)13 merges (Hugging Face model tree)9 adapters (Hugging Face model tree)97 quantized40 fine-tunes13 merges9 adapters
97 quantized (Hugging Face model tree)40 fine-tunes (Hugging Face model tree)13 merges (Hugging Face model tree)9 adapters (Hugging Face model tree)97 quantized40 fine-tunes13 merges9 adapters
GGUF OBLITERATUS/Ornith-1.5-9B-OBLITERATED for llama.cpp, Ollama, LM Studio 749.4k downloads MLX junafinity/Ornith-1.5-9B-uncensored-MLX-8bit for Apple silicon 4.1k downloads INT4 letechlead/Ornith-1.5-9B-INT4-W4A16-AutoRound for vLLM, SGLang 15.7k downloads

All variants →

Reference

Everything else, folded: the full tables, every source.

Benchmarks

8 rows

Benchmarks 8 rows

measured third-party leaderboard rowauthor-reported number from the model card or eval files (8)No confidence intervals unless the source gave them; settings differ between rows, so compare within a benchmark, not across. Date = the benchmark's own date when it gives one, else when we read it.
BenchmarkValueSetting / notesReported byDate
author-reported GPQA Diamond · Reason through hard problems86.4—model author2026-10-07
author-reported MCP-Atlas54.2—model author2026-10-07
author-reported NL2Repo · Write code with agents32.4—model author2026-10-07
author-reported SWE Atlas - QnA20.6—model author2026-10-07
author-reported SWE-bench Multilingual · Write code with agents54.4—model author2026-10-07
author-reported SWE-bench Pro · Write code with agents47.5—model author2026-10-07
author-reported SWE-bench Verified · Write code with agents70.6—model author2026-10-07
author-reported Toolathlon · Call tools and APIs41.2—model author2026-10-07

What the card says

1 cited sentence

Summary generated, cited

Ornith-1.5-9B is a 9.65B parameter dense model designed for text generation tasks.[1][2]

  1. parameters human
  2. “text-generation”

Generated 2026-10-07 15:13 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.

Use-case evidence

7 use cases

What it is good at 7 use cases

Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.

Facts

20 fields

Listed in LLMs

Facts

Pipeline tagtext-generation
Librarytransformers
ArchitecturesQwen3_5ForConditionalGeneration
Model typeqwen3_5
Dtypebfloat16
Params by dtypeBF16: 9,653,104,368
Size bandL
MoEno
Context length262,144 (config.json max_position_embeddings)
LicenseMIT (mit)
Hub snapshot2026-10-25

Servability

vLLM · SGLang

Servability

vLLM: supported · Qwen3_5ForConditionalGeneration · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen3_5ForConditionalGeneration · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed

vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.

Hardware fit

3 estimates

Hardware fit ESTIMATE

Weights-only memory for Ornith-1.5-9B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.

BF16 (2 bytes/param)23 GB · Single GPU (<24 GB)
FP8 (1 byte/param)12 GB · Single GPU (<24 GB)
INT4 (0.5 bytes/param)5.8 GB · Single GPU (<24 GB)

Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.

All variants

40 variants

Lineage

Original model Family: ornith-ai/Ornith-1.5 159 derivatives on the Hub

quantized: 97finetune: 40merge: 13adapter: 9

Top variants (40 shown, ranked by downloads)

Relation
Quant
Size
40 rows
Variant Relation Quant Params Band License Author Likes Downloads
OBLITERATUS/Ornith-1.5-9B-OBLITERATED quantized GGUF 9.65B L mit OBLITERATUS 86 749.4k
dealignai/Ornith-1.5-9B-UNCENSORED-GGUF quantized GGUF — — mit dealignai 31 638.8k
ornith-ai/Ornith-1.5-9B-DFlash finetune — 1.29B S mit ornith-ai 25 319k
AtomicChat/Ornith-1.5-9B-GGUF quantized GGUF — — mit AtomicChat 78 123.9k
bartowski/Ornith-1.5-9B-GGUF quantized GGUF — — mit bartowski 14 56.2k
protoLabsAI/Ornith-1.5-9B-MTP-GGUF quantized GGUF — — mit protoLabsAI 42 47.9k
Thunder13240/Ornith-1.5-9B-heretic-GGUF quantized GGUF — — mit Thunder13240 24 36.9k
letechlead/Ornith-1.5-9B-INT4-W4A16-AutoRound quantized INT4 4.34B M mit letechlead 0 15.7k
guell00/VELUM-Coder quantized GGUF — — mit guell00 11 15k
cyankiwi/Ornith-1.5-9B-AWQ-INT4 quantized AWQ 9.41B L mit cyankiwi 3 13.4k
zaakirio/Ornith-1.5-9B-Uncensored-GGUF quantized GGUF — — mit zaakirio 5 6.9k
Soulfate24/Ornith-1.5-9B-MTP-DFlash-Paretrix quantized GGUF — — mit Soulfate24 0 6.8k
audreyt/Ornith-1.5-9B-DFlash-GGUF quantized GGUF — — apache-2.0 audreyt 3 6.5k
emperorofrome/Gmcoder merge GGUF 9.41B L mit emperorofrome 25 5.3k
junafinity/Ornith-1.5-9B-uncensored-GGUF-8bit quantized GGUF — — apache-2.0 junafinity 9 5.9k
ressl/Ornith-1.5-9B-uncensored-GGUF finetune GGUF — — mit ressl 2 6k
NANI-Nithin/Ornith-1.5-9B-GGUF quantized GGUF — — apache-2.0 NANI-Nithin 1 5.8k
huihui-ai/Huihui-Ornith-1.5-9B-abliterated finetune — 9.41B L mit huihui-ai 39 3.8k
mradermacher/Ornith-1.5-9B-i1-GGUF quantized GGUF — — mit mradermacher 2 5.1k
jashepp/Ornith-1.5-9B-MXFP4_Hybrid-Imatrix-GGUF quantized Imatrix — — mit jashepp 2 4.3k
junafinity/Ornith-1.5-9B-uncensored-MLX-8bit quantized MLX 9.41B L apache-2.0 junafinity 4 4.1k
kingjones777/Ornith-1.5-9B-ROCmFPX-AGENT-GGUF quantized GGUF — — apache-2.0 kingjones777 0 4.2k
PocketAiHub/Ornith-1.5-9B-Abliterated-GGUF quantized GGUF — — mit PocketAiHub 3 3.7k
mradermacher/Ornith-1.5-9B-GGUF quantized GGUF — — mit mradermacher 3 3.6k
prithivMLmods/Ornith-1.5-9B-GGUF quantized GGUF — — mit prithivMLmods 2 3.4k
Shiftedx/ornith-1.5-9b-attention8-bf16recurrence-vision-mlx quantized mlx 9.41B L mit Shiftedx 0 3.3k
tiyuvta/Ornith-1.5-9B-NVFP4-MTP-GGUF quantized NVFP4 — — mit tiyuvta 4 3k
Sohailhosseini/Ornith-1.5-9B-AWQ-W4A16 quantized AWQ 9.41B L mit Sohailhosseini 0 3.2k
wepiqx/Ornith-1.5-9B-MTP-ASHQ1-GGUF quantized GGUF — — apache-2.0 wepiqx 1 2.9k
protoLabsAI/Ornith-1.5-9B-NVFP4 quantized NVFP4 9.41B L mit protoLabsAI 0 2.9k
kizzlah/Ornith-1.5-9B-GGUF quantized GGUF — — mit kizzlah 0 2.7k
DarkKitsune/Qwen3.5-9B-Qworus-V2 merge GGUF 9.41B L — DarkKitsune 4 2.5k
junafinity/Ornith-1.5-9B-uncensored finetune — 9.41B L apache-2.0 junafinity 14 1.9k
vonvonhero/Ornith-1.5-9B-Uncensored-GGUF quantized GGUF — — mit vonvonhero 1 2.2k
OliviaRossi/MiMo-Ornith-9B-AGSI merge — 9.65B L apache-2.0 OliviaRossi 12 1.7k
mistrjirka/Ornith-1.5-9B-PXQ-GGUF quantized GGUF — — mit mistrjirka 0 2.3k
0xSojalSec/Ornith-1.5-9B-OBLITERATED quantized GGUF 9.65B L mit 0xSojalSec 0 2.2k
minte1431/Ornith-1.5-9B-OBLITERATED quantized GGUF 9.65B L mit minte1431 0 2.1k
xoykor/Ornith-1.5-9B-DSH-Agentic-GPT-5.6-Sol-Distill adapter GGUF — — mit xoykor 9 1.6k
ressl/Ornith-1.5-9B-uncensored-MLX-6bit finetune MLX 8.95B M mit ressl 3 1.9k

Repository changelog

29 commits
changed in the last 7 daysOther: 2026-10-24, 2026-10-24all changes

Repository changelog 29 commits

DateCommitSHA
2026-08-23 Upload model-00004-of-00004.safetensors with huggingface_hub 489cb97
2026-08-23 Upload model-00004-of-00004.safetensors with huggingface_hub 8347866
2026-08-23 Upload model.safetensors.index.json with huggingface_hub ae092ce
2026-08-20 Update README.md c927ad7
2026-08-19 Update README.md 98db59b
2026-08-19 Update README.md e7509fb
2026-08-19 Update README.md fb0fe62
2026-08-19 Update README.md 53979d3
2026-08-19 Update README.md 362bd07
2026-08-19 Update README.md 128348e
2026-08-19 Update README.md 474e796
2026-08-19 Update README.md f176c83
2026-08-19 Update README.md 4762960
2026-08-19 Update README.md 4bd542b
2026-08-19 Update README.md fad58dd
2026-08-19 Upload ornith_9b_eval.png 72d4efa
2026-08-19 Delete assets/ornith_9b_eval.png df115da
2026-08-19 Upload ornith_9b_eval.png d1ec1b2
2026-08-19 Delete assets/ornith_9b_eval.png a5e4dd8
2026-08-19 Update README.md b3353fc
2026-08-19 Update README.md bec2320
2026-08-19 Upload ornith_9b_eval.png 1a5025e
2026-08-19 Delete assets/ornith_9b_eval.png 1edf7af
2026-08-19 Upload 2 files 5e95215
2026-08-19 Update README.md 7c0acf9
2026-08-19 Update README.md aadd203
2026-08-18 Delete latest_checkpointed_iteration.txt with huggingface_hub 40edaaf
2026-08-18 Upload folder using huggingface_hub 9f70491
2026-08-18 initial commit 4d854e9

All providers

1 provider · 1 endpoint

Where to run 1 providers · 1 endpoints

Hosted endpoints observed for Ornith-1.5-9B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.

ProviderModel refContextQuant$/M in$/M outtok/sTTFTToolsStatusObserved
Featherless
hf
 ornith-ai/Ornith-1.5-9B — — — — — — — live 2026-10-25

Articles and coverage

1 piece

Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.

Official1

Ornith-1.5: From Self-Scaffolding to Self-Improvement
ornith.ai2026-08-19officialHN front pageabout the Ornith 1.5 familyHN 216 pts · 76 comments2,250 words
Ornith-1.5 extends its self-scaffolding framework into a complete self-improvement loop that jointly optimizes task generation, scaffold construction, and solution rollouts. The system uses GRPO to optimize the loop based on validity, frontier difficulty, and novelty rewards, with a target frontier difficulty of 0.2. The 397B MoE variant scores 86.1 on Terminal-Bench 2.1 and 56.0 on DeepSWE, matching Claude Opus 4.8 while outperforming peers like GLM-5.2.written, reviewed, cited
  • The 9B model can be deployed on iPhone and Android devices while outperforming larger models like Gemma 4-31B.
  • The system proposes progressively harder tasks that go beyond what the model has already solved to expose capability gaps.
  • Validity is treated as a hard gate to prevent malformed tasks from receiving reward simply because they appear difficult.

Similar models

3 models

Similar models

  • Llama-3.1-8B-Instruct · 8.03B
    same task (text-generation), more providers
  • gemma-2-9b-it · 9.24B
    same size band (L), same task (text-generation), more providers
  • Qwen3-8B · 8.19B
    same task (text-generation), more providers, permissive license

Sources

3 sources

Sources

Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:41 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:13 UTC · Last verified 2026-10-25

Follow Ornith-1.5-9B

Watch this model

Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.