LLM · Turintech

Qwen

Eine Modellfamilie als lebendes Objekt: Varianten, ausgewählte Leistungsdaten und alle verbundenen BrunoSan-Quellensignale.

Erwähnungen

in 30 Tagen

Quellen

5Varianten

Im Universe erkunden →History öffnen →

Modelle

Alle bekannten Varianten dieser Familie. Ein direkter Artikel-Link springt exakt zur genannten Variante.

Turintech · Qwen

Qwen3.5-4B-Artemis-GGUF

KNOWN

Technische Leistungsdaten sind für diese Variante noch nicht vollständig bestätigt.

—Kontext

—Max. Output

—Input

—Output

—Input / 1M

—Output / 1M

—Tools

—Reasoning

—Structured

HF · Lizenz apache-2.0 Erwähnungen →Universe →Hugging Face ↗

Benchmarks

Benchmark	Wert	Beleg
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, protocol m3max-2026-10-15, text wiki (lower is closer)	0.0106	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, protocol m3max-2026-10-15, text wiki (lower is closer)	0.0178	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0126	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, protocol m3max-2026-10-15, text chat (lower is closer)	0.0316	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol rtx5090-2026-10-23, text chat (higher is closer)	95.7	Quelle ↗abgerufen 2026-10-25

Weitere 22 Ergebnisse

Benchmark	Wert	Beleg
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q5_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.010497	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol m3max-2026-10-15, text wiki (higher is closer)	95	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-23, text chat (higher is closer)	96.2	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q4_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.021046	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, protocol m3max-2026-10-15, text chat (lower is closer)	0.0218	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, protocol m3max-2026-10-15, text chat (lower is closer)	0.0179	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol m3max-2026-10-15, text chat (higher is closer)	94.8	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0191	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q5_K_S, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.0125	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0144	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol m3max-2026-10-15, text wiki (higher is closer)	95.6	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0076	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text wiki (higher is closer)	93.5	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol rtx5090-2026-10-23, text wiki (higher is closer)	94.4	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-23, text wiki (higher is closer)	95.1	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0094	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, protocol m3max-2026-10-15, text wiki (lower is closer)	0.0079	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text chat (higher is closer)	94.9	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0097	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol m3max-2026-10-15, text wiki (higher is closer)	93.7	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol m3max-2026-10-15, text chat (higher is closer)	96.1	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol m3max-2026-10-15, text chat (higher is closer)	95.5	Quelle ↗abgerufen 2026-10-25

Turintech · Qwen

Qwen3.5-0.8B-Artemis-GGUF

KNOWN

Technische Leistungsdaten sind für diese Variante noch nicht vollständig bestätigt.

—Kontext

—Max. Output

—Input

—Output

—Input / 1M

—Output / 1M

—Tools

—Reasoning

—Structured

HF · Lizenz apache-2.0 Erwähnungen →Universe →Hugging Face ↗

Benchmarks

Benchmark	Wert	Beleg
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol cpu-replay-2026-10-16, text chat (higher is closer)	93.353	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-20, text wiki (higher is closer)	94.2	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q5_K_M, nats, protocol cpu-replay-2026-10-16, text wiki (lower is closer)	0.058802	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol cpu-replay-2026-10-16, text chat (lower is closer)	0.011123	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol cpu-replay-2026-10-16, text wiki (lower is closer)	0.017288	Quelle ↗abgerufen 2026-10-25

Weitere 17 Ergebnisse

Benchmark	Wert	Beleg
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-20, text chat (lower is closer)	0.0163	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-20, text chat (lower is closer)	0.0092	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q5_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.013993	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-20, text wiki (higher is closer)	92.7	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-20, text chat (higher is closer)	93.6	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q4_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.022265	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-20, text chat (higher is closer)	95.1	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-20, text wiki (lower is closer)	0.0098	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q5_K_M, nats, protocol cpu-replay-2026-10-16, text chat (lower is closer)	0.063854	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol cpu-replay-2026-10-16, text chat (higher is closer)	94.719	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q4_K_M, nats, protocol cpu-replay-2026-10-16, text wiki (lower is closer)	0.083643	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q4_K_M, nats, protocol cpu-replay-2026-10-16, text chat (lower is closer)	0.108646	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-20, text wiki (lower is closer)	0.0165	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol cpu-replay-2026-10-16, text wiki (higher is closer)	92.523	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol cpu-replay-2026-10-16, text wiki (lower is closer)	0.012005	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol cpu-replay-2026-10-16, text chat (lower is closer)	0.017126	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol cpu-replay-2026-10-16, text wiki (higher is closer)	93.667	Quelle ↗abgerufen 2026-10-25

Turintech · Qwen

Qwen3.5-2B-Artemis-GGUF

KNOWN

Technische Leistungsdaten sind für diese Variante noch nicht vollständig bestätigt.

—Kontext

—Max. Output

—Input

—Output

—Input / 1M

—Output / 1M

—Tools

—Reasoning

—Structured

HF · Lizenz apache-2.0 Erwähnungen →Universe →Hugging Face ↗

Benchmarks

Benchmark	Wert	Beleg
C4 validation, held-out half (chunks: 40; context (tokens): 1024)Mean KLD vs BF16, Q5_K_M, protocol gtx1080ti-2026-10-15, text c4 (lower is closer)	0.007	Quelle ↗abgerufen 2026-10-25
C4 validation, held-out half (chunks: 40; context (tokens): 1024)Mean KLD vs BF16, Q4_K_M, protocol gtx1080ti-2026-10-15, text c4 (lower is closer)	0.0119	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-23, text wiki (higher is closer)	95	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q5_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.010963	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0068	Quelle ↗abgerufen 2026-10-25

Weitere 9 Ergebnisse

Benchmark	Wert	Beleg
OPUS-100 test, held-out half (chunks: 30; context (tokens): 1024)Mean KLD vs BF16, Q5_K_M, protocol gtx1080ti-2026-10-15, text opus (lower is closer)	0.008	Quelle ↗abgerufen 2026-10-25
OPUS-100 test, held-out half (chunks: 30; context (tokens): 1024)Mean KLD vs BF16, Q4_K_M, protocol gtx1080ti-2026-10-15, text opus (lower is closer)	0.0137	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0121	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q4_K_M, nats, protocol cpu-probe-2026-10-15, text probe6 (lower is closer)	0.016292	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_M, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0073	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_M, protocol rtx5090-2026-10-23, text chat (higher is closer)	95.8	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text wiki (higher is closer)	93.8	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0125	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text chat (higher is closer)	94.6	Quelle ↗abgerufen 2026-10-25

Turintech · Qwen

Qwen3.5-9B-Artemis-GGUF

KNOWN

Technische Leistungsdaten sind für diese Variante noch nicht vollständig bestätigt.

—Kontext

—Max. Output

—Input

—Output

—Input / 1M

—Output / 1M

—Tools

—Reasoning

—Structured

HF · Lizenz apache-2.0 Erwähnungen →Universe →Hugging Face ↗

Benchmarks

Benchmark	Wert	Beleg
WikiText-2 test (chunks: 60; context (tokens): 512)PPL ratio vs BF16, IQ4_XS, protocol gb10-replay-2026-10-15, text wiki (1 is the reference's perplexity)	1.007853	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q5_K_S, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.036	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q4_K_M, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.115	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol gb10-replay-2026-10-15, text wiki (higher is closer)	95.1	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.0089	Quelle ↗abgerufen 2026-10-25

Weitere 34 Ergebnisse

Benchmark	Wert	Beleg
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text chat (higher is closer)	96.1	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, IQ4_XS, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0138	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.0093	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q5_K_S, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.062	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0095	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)PPL ratio vs BF16, Q4_K_M, protocol gb10-replay-2026-10-15, text wiki (1 is the reference's perplexity)	1.012493	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, IQ4_XS, protocol rtx5090-2026-10-23, text wiki (higher is closer)	93.2	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, IQ4_XS, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.0207	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, IQ4_XS, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.098	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, IQ4_XS, protocol gb10-replay-2026-10-15, text wiki (higher is closer)	93.3	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)PPL ratio vs BF16, Q5_K_S, protocol gb10-replay-2026-10-15, text wiki (1 is the reference's perplexity)	1.011519	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol rtx5090-2026-10-23, text wiki (higher is closer)	94	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.0149	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, Q4_K_M, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.063	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, IQ4_XS, protocol gb10-replay-2026-10-15, text chat (higher is closer)	95	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, IQ4_XS, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0211	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, IQ4_XS, nats, protocol gb10-probe-2026-10-15, text probe6 (lower is closer)	0.022671	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)PPL ratio vs BF16, Q5_K_S, protocol gb10-replay-2026-10-15, text chat (1 is the reference's perplexity)	1.004657	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, IQ4_XS, protocol rtx5090-2026-10-23, text chat (higher is closer)	94.8	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.0058	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol rtx5090-2026-10-23, text chat (higher is closer)	96.6	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol gb10-replay-2026-10-15, text wiki (higher is closer)	94.1	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)PPL ratio vs BF16, IQ4_XS, protocol gb10-replay-2026-10-15, text chat (1 is the reference's perplexity)	1.006549	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)99th-percentile KLD vs BF16, IQ4_XS, nats, protocol gb10-replay-2026-10-15, text wiki (lower is closer)	0.17	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q4_K_M, protocol gb10-replay-2026-10-15, text chat (higher is closer)	96	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol gb10-replay-2026-10-15, text chat (higher is closer)	96.7	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, IQ4_XS, nats, protocol gb10-replay-2026-10-15, text chat (lower is closer)	0.0137	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)PPL ratio vs BF16, Q4_K_M, protocol gb10-replay-2026-10-15, text chat (1 is the reference's perplexity)	1.005059	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text wiki (lower is closer)	0.0145	Quelle ↗abgerufen 2026-10-25
WikiText-2 test (chunks: 60; context (tokens): 512)Same top p vs BF16, Q5_K_S, protocol rtx5090-2026-10-23, text wiki (higher is closer)	95.3	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q4_K_M, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0088	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q5_K_S, nats, protocol gb10-probe-2026-10-15, text probe6 (lower is closer)	0.009184	Quelle ↗abgerufen 2026-10-25
Six fixed documents: two recent Wikipedia extracts, two authored code files and two authored chats (chunks: 2 per document; context (tokens): 256)Mean KLD vs BF16, Q4_K_M, nats, protocol gb10-probe-2026-10-15, text probe6 (lower is closer)	0.021976	Quelle ↗abgerufen 2026-10-25
OpenAssistant validation conversations (English threads as User/Assistant text) (chunks: 60; context (tokens): 512)Mean KLD vs BF16, Q5_K_S, nats, protocol rtx5090-2026-10-23, text chat (lower is closer)	0.0059	Quelle ↗abgerufen 2026-10-25

Turintech · Qwen

Qwen3.8-27B-Artemis-GGUF

KNOWN

Technische Leistungsdaten sind für diese Variante noch nicht vollständig bestätigt.

—Kontext

—Max. Output

—Input

—Output

—Input / 1M

—Output / 1M

—Tools

—Reasoning

—Structured

HF · Lizenz apache-2.0 Erwähnungen →Universe →Hugging Face ↗

Benchmarks

Benchmark	Wert	Beleg
Long documents (chunks: 50 (one per text); context (tokens): 2048)Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text long_document (lower is closer)	1.095	Quelle ↗abgerufen 2026-10-25
Coding (chunks: 50 (one per text); context (tokens): 256)Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text coding (lower is closer)	1.514	Quelle ↗abgerufen 2026-10-25
Tool text (BFCL conversations, not the scored tasks) (chunks: 50 (one per text); context (tokens): 256)Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text tooltext (higher is closer)	66.5	Quelle ↗abgerufen 2026-10-25
All 250 texts, each text weighted equally (chunks: 250 (one per text); context (tokens): 256 (2048 for long documents))Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text all (higher is closer)	68.6	Quelle ↗abgerufen 2026-10-25
Long documents (chunks: 50 (one per text); context (tokens): 2048)Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text long_document (higher is closer)	67.6	Quelle ↗abgerufen 2026-10-25

Weitere 7 Ergebnisse

Benchmark	Wert	Beleg
Multilingual (chunks: 50 (one per text); context (tokens): 256)Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text multilingual (higher is closer)	69.1	Quelle ↗abgerufen 2026-10-25
Chat (chunks: 50 (one per text); context (tokens): 256)Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text chat (lower is closer)	0.928	Quelle ↗abgerufen 2026-10-25
Multilingual (chunks: 50 (one per text); context (tokens): 256)Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text multilingual (lower is closer)	0.694	Quelle ↗abgerufen 2026-10-25
Tool text (BFCL conversations, not the scored tasks) (chunks: 50 (one per text); context (tokens): 256)Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text tooltext (lower is closer)	2.723	Quelle ↗abgerufen 2026-10-25
All 250 texts, each text weighted equally (chunks: 250 (one per text); context (tokens): 256 (2048 for long documents))Mean KLD vs BF16, IQ2_XXS, nats, protocol rtx5090-2026-10-25, text all (lower is closer)	1.391	Quelle ↗abgerufen 2026-10-25
Coding (chunks: 50 (one per text); context (tokens): 256)Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text coding (higher is closer)	67.6	Quelle ↗abgerufen 2026-10-25
Chat (chunks: 50 (one per text); context (tokens): 256)Same top p vs BF16, IQ2_XXS, protocol rtx5090-2026-10-25, text chat (higher is closer)	72.1	Quelle ↗abgerufen 2026-10-25

BrunoSan Chronik

Alle verbundenen Artikel, neueste zuerst.

Noch keine verbundenen Artikel.

Modelldaten und Nachrichten aus BrunoSan AI-News, OpenRouter und Hugging Face. Benchmark-Werte stammen aus den jeweils verlinkten Quellen; Testbedingungen können abweichen. Modelle bleiben auch ohne Benchmark-Werte sichtbar.