LFM2.5-2.6B
LFM2.5-2.6B is Liquid AI's open-weight 2.6B dense hybrid text→text model (LFM2 architecture) post-trained for on-device agentic workloads: planning, native tool calling, and multi-step tasks. ~2.69B parameters, trained on ~34T tokens, 128K context (131,072 tokens), 128K vocabulary. Always thinks before answering. Day-one support for Transformers, vLLM, SGLang, llama.cpp, MLX, and ONNX. Hugging Face: LiquidAI/LFM2.5-2.6B (+ Base). License: LFM Open License v1.0 (lfm1.0). Catalog benches include AIME25, LiveCodeBench v6, IFBench, Multi-IF, BFCL-v4, PinchBench, Claw-Eval, τ³-Banking 0.0567 (Liquid harness; scale differs from some AA τ³ rows), and aa-omniscience-index −29.50 (signed). No catalog id for ToolSandbox 77.83, BrowseComp+ (OpenClaw) 26.89, or IFStruct 85.49 — recorded here only. Sources: https://www.liquid.ai/blog/lfm2-5-2-6b · https://huggingface.co/LiquidAI/LFM2.5-2.6B. No official hosted provider row in this catalog (no sibling Liquid/OpenRouter provider pattern present); OpenRouter free route liquid/lfm-2.5-2.6b:free exists externally but was not added.
Identité
| Organisation | Liquid AI |
|---|---|
| Pays | — |
| Famille | — |
| Type | llm |
| Date de sortie | 2026-08-04 |
| Licence | lfm open license v1.0 |
| Licence commerciale | oui |
| Licence open source | non |
| Poids ouverts | oui |
| Auto-hébergement (DGX Spark) | oui — ≤ 107 Md (Q8/FP8) |
| Modalités | — |
| Fenêtre de contexte | 131 072 |
| KV cache / token (Q8) | 30.7 ko/token |
| Paramètres totaux | 2.69 B |
| Paramètres actifs (session) | 2.7 B |
| Source des paramètres actifs | actifs = total (modèle dense, config HF (LiquidAI/LFM2.5-2.6B)) ; total : config Hugging Face (LiquidAI/LFM2.5-2.6B) |
| Curation des paramètres actifs | 2026-11-04 |
| Prix par session | — |
| Tokens / tâche (Data Players) | 1.2 M |
| Tokens / tâche (Terminal-Bench) | 83.3 M (est.) |
| Ratio Terminal-Bench / Data Players | 69.4× |
| Coût / tâche (Terminal-Bench) | — |
| Tokens d'entraînement | 34 000 000 000 000 |
| Coupure de connaissance | — |
| Créé le | 2026-08-12T15:38:35.392575Z |
| Mis à jour le | 2026-11-04T06:04:49.333512Z |
| Source | LLM Stats |
| URL | https://llm-stats.com/models/lfm-2.5-2.6b |
Prix par fournisseur
Statut du modèle (offres + auto-hébergement) : partiel — auto-hébergement DGX Spark (≤ 107 Md).
Toutes sources confondues : catalogue llm-stats, routeurs EUrouter/Cortecs et catalogues directs (OVHcloud, Scaleway, Infomaniak). Les prix sont convertis en USD pour la comparaison ; le prix publié et le taux BCE utilisé sont dans l'infobulle. Les prix inconnus restent indiqués « — ».
Aucune donnée.
Souveraineté qualifiée par offre (modèle × fournisseur) ; survolez le badge pour les critères.
SouverainetéScores par catégorie (top_scores)
| Catégorie | Score |
|---|---|
| chat | 0.80 |
| instruction_following | 0.70 |
| knowledge | -29.50 |
| language | 0.80 |
| math | 0.50 |
| reasoning | -5.50 |
| science | -29.50 |
| structured_output | 0.80 |
| general | -2.80 |
| agents | 0.50 |
| code | 0.70 |
| communication | 0.80 |
| tool_calling | 0.30 |
Scores par benchmark
| Benchmark | Catégorie | Score | Max | Normalisé (API) | Vérifié | Auto-déclaré | Daté du |
|---|---|---|---|---|---|---|---|
| aa-omniscience-index | knowledge | -29.50 | 200.00 | — | non | oui | 2026-11-04T06:04:49.403082Z |
| aime-2025 | math | 0.52 | 1.00 | 0.52 | non | oui | 2026-11-04T06:04:49.344532Z |
| bfcl-v4 | tool_calling | 0.57 | 1.00 | 0.57 | non | oui | 2026-11-04T06:04:49.370789Z |
| claw-eval | code | 0.63 | 1.00 | 0.63 | non | oui | 2026-11-04T06:04:49.413173Z |
| ifbench | instruction_following | 0.59 | 1.00 | 0.59 | non | oui | 2026-11-04T06:04:49.362064Z |
| livecodebench-v6 | reasoning | 0.59 | 1.00 | 0.59 | non | oui | 2026-11-04T06:04:49.353145Z |
| multi-if | chat | 0.80 | 1.00 | 0.80 | non | oui | 2026-11-04T06:04:49.377086Z |
| pinchbench | code | 0.68 | 1.00 | 0.68 | non | oui | 2026-11-04T06:04:49.374012Z |
| tau3-banking | agents | 0.06 | 1.00 | 0.06 | non | oui | 2026-11-04T06:04:49.393372Z |
Classements
| Catégorie | Rang | Score | Note conservatrice | Benchmarks évalués | Méthode | Daté du | Provenance |
|---|---|---|---|---|---|---|---|
| chat | 44 | 0.80 | 21.09 | 1 | trueskill | 2026-11-04T08:08:43.789169Z | rankings |
| knowledge | 36 | -29.50 | -2.80 | 1 | trueskill | 2026-11-04T08:08:45.139626Z | rankings |
| agents | 150 | — | 7.61 | 4 | trueskill | 2026-10-19 | model_history |
| code | 136 | — | 16.91 | 2 | trueskill | 2026-10-19 | model_history |
Données
Dates de curation (récupération depuis l'API) : models 2026-11-04 · benchmarks 2026-11-04 · scores 2026-11-04 · updates 2026-11-04 · account 2026-11-04 · model_history 2026-11-04 · rankings 2026-11-04