Skip to content
AI Atlas
Time machine

AI Atlas as of 11 Sept 2026

Pick any date: the models that existed, their context lengths and status at the time, the provider prices valid that day, the benchmark leaders known by then, the hardware. Every value opens the claim that establishes it, with its validity interval.

ObservedThis date is inside the observation history: values are claims that were current that day.

Observation history starts . AI Atlas observation history starts at 2026-09-11T20:50:56.012288+00:00. The state is taken from claims, prices and results as they were known at the end of that UTC day.

Models existingCanonical models whose release date (or first dated claim) is on or before the date.

1,515

Offers validPrice rows whose validity interval covers the date (append-only history).

617

Benchmark leadersBenchmarks with at least one result observed by the date.

23

HardwareHardware entities released on or before the date.

25

BasisReconstructed = before the first observation; observed = inside the recorded history.

observed

Models

Models as of 11 Sept 2026 1,515

All →
Models as of 2026-09-11
ModelOrganizationParamsContext (as of)Context (today)OpennessStatus (as of)Status (today)ReleasedBasisHistory
Fugu MaxSakana—same——active11 Sept 2026observedas of 2026-09-11 →
Fugu Ultra v2Sakana—same——active11 Sept 2026observedas of 2026-09-11 →
DeepSeek-V4.1-FlashDeepSeek—sameOpen weights—active10 Sept 2026observedas of 2026-09-11 →
Ling 3.0 Flash VLinclusionAI—131.1KOpen weights—active10 Sept 2026observedas of 2026-09-11 →
North Small TranslateCoheresameOpen weightsAnnouncedsame10 Sept 2026observedas of 2026-09-11 →
Mercury 2.5Inception—same——active8 Sept 2026observedas of 2026-09-11 →
GPT-6 Astra ProOpenAI—same——active4 Sept 2026observedas of 2026-09-11 →
gpt-6-astraOpenAI—sameProprietary—active4 Sept 2026observedas of 2026-09-11 →
Qwen3.8 Max (0902)Qwen—same——active3 Sept 2026observedas of 2026-09-11 →
Gemini 3.8 FlashGoogle—sameProprietaryActivesame2 Sept 2026observedas of 2026-09-11 →
Muse Spark 1.3Meta AI—1.05MProprietary—active2 Sept 2026observedas of 2026-09-11 →
Muse Spark 1.3 ContributorMeta AI—same——active2 Sept 2026observedas of 2026-09-11 →
Claude Fable 5.1Anthropic—sameProprietaryActivesame1 Sept 2026observedas of 2026-09-11 →
DeepSeek V4 Flash Vision ExpDeepSeeksameProprietaryPreviewsame31 Aug 2026observedas of 2026-09-11 →
Granite 4.2 8BIBM—sameOpen weights—active31 Aug 2026observedas of 2026-09-11 →
Hy4 previewTencent—same——active28 Aug 2026observedas of 2026-09-11 →
Ling 3.0 Flash FininclusionAI—same——active27 Aug 2026observedas of 2026-09-11 →
Qwen3.8 FlashQwen—1MOpen weights—active26 Aug 2026observedas of 2026-09-11 →
GLM 5.3Z.ai (Zhipu AI)1.31MOpen weights—active25 Aug 2026observedas of 2026-09-11 →
GLM 5.3 FlashZ.ai (Zhipu AI)1.31MOpen weights—active25 Aug 2026observedas of 2026-09-11 →
pplx-computer-qwen-3-8-27b-dflash2-20260824Perplexity AI——Open weights—active25 Aug 2026observedas of 2026-09-11 →
WeMM-Embedding-2BTencent——Open weights—active25 Aug 2026observedas of 2026-09-11 →
Muse Spark 1.2 ContributorMeta AI—same——active21 Aug 2026observedas of 2026-09-11 →
pplx-computer-vllm-dflash2Perplexity AI———Open weights—active20 Aug 2026observedas of 2026-09-11 →
pplx-qwen-3-8-27b-dflash2-20260819Perplexity AI——Open weights—active19 Aug 2026observedas of 2026-09-11 →

Showing 25 of 1,515 models that existed on this date (API cap 25). Open a model's History tab for its full as-of record.

“Today” columns compare the as-of value with the current record; “same” means no recorded change. Click any as-of value for the claim behind it (source, tier, valid from → to).

Provider prices

Offers valid on 11 Sept 2026 617

All →
Provider prices valid on 2026-09-11
ModelProviderInput / 1MOutput / 1MContextValid fromValid toSource
Ternary Bonsai 27BTogether AI———11 Sept 2026currentT1source
Inkling SmallOpenRouter——1.05M11 Sept 2026currentT2source
Laguna XS 2.1OpenRouter——262.1K11 Sept 2026currentT2source
Free Models RouterOpenRouter——200K11 Sept 2026currentT2source
Nex-N2.5-Mini (free)OpenRouter——262.1K11 Sept 2026currentT2source
LFM2.5-2.6B (free)OpenRouter——65.5K11 Sept 2026currentT2source
Ling 3.0 Flash FinOpenRouter——262.1K11 Sept 2026currentT2source
InklingOpenRouter——1.05M11 Sept 2026currentT2source
Lyria 3Google Gemini API——1.05M11 Sept 2026currentT2source
Nex-N2.5-Pro (free)OpenRouter——262.1K11 Sept 2026currentT2source
Nemotron 3 SuperNVIDIA NIM / build.nvidia.com——262.1K11 Sept 2026currentT2source
Nemotron 3.5 Content SafetyNVIDIA NIM / build.nvidia.com——128K11 Sept 2026currentT2source
North Mini Code (free)Cohere API——256K11 Sept 2026currentT2source
Gemma 4 26B A4BGoogle Gemini API——262.1K11 Sept 2026currentT2source
Nemotron 3 Nano Omni (free)NVIDIA NIM / build.nvidia.com——256K11 Sept 2026currentT2source
Dots3-Note Preview (free)OpenRouter——512K11 Sept 2026currentT2source
Ling 3.0 Flash VLOpenRouter——262.1K11 Sept 2026currentT2source
Leanstral 1.5Mistral AI La Plateforme———11 Sept 2026currentT1source
Laguna S 2.1OpenRouter——262.1K11 Sept 2026currentT2source
Gemma 4 31BGoogle Gemini API——262.1K11 Sept 2026currentT2source
NVIDIA Nemotron 3.5 Lightning 30B A3BNVIDIA NIM / build.nvidia.com——1M11 Sept 2026currentT2source
Lyria 3 Pro PreviewGoogle Gemini API——1.05M11 Sept 2026currentT2source
Ling 3.0 Flash Sante (free)OpenRouter——262.1K11 Sept 2026currentT2source
Nemotron 3 UltraNVIDIA NIM / build.nvidia.com——1M11 Sept 2026currentT2source
Llama Prompt Guard 2 22MGroqCloud——51211 Sept 2026currentT1source

Showing 25 of 617 offers valid on this date.

offers whose validity interval covers the date (price rows are append-only)

Benchmark leaders

Leaders known by 11 Sept 2026 23

Benchmark leaders as of 2026-09-11
BenchmarkLeaderOrganizationScoreMetric · groupTrustModels ranked
Aider polyglotgpt-5OpenAI86.7pass_rate_2official-benchmark61
Aider polyglot — well-formed responsesCodestral 25.01Mistral AI100percent_cases_well_formedofficial-benchmark61
Artificial Analysis Intelligence IndexClaude Fable 5.1Anthropic53.37indexindependent-evaluator477
GPQA Diamondgpt-6-astraOpenAI96.26accuracyaccuracy · variant=GPQA Diamond · evaluator=Artificial Analysisindependent-evaluator459
Humanity's Last ExamClaude Fable 5.1Anthropic59.13accuracyaccuracy · evaluator=Artificial Analysisindependent-evaluator453
IFBenchGrok 4.3xAI83.33accuracyaccuracy · evaluator=Artificial Analysisindependent-evaluator333
LiveBenchClaude Fable 5.1Anthropic83.41global_averageofficial-benchmark57
LiveBench Agentic CodingDeepSeek-V4.1-FlashDeepSeek77.27average scoreaverage score · variant=Agentic Codingofficial-benchmark57
LiveBench CodingClaude Fable 5.1Anthropic86.38average scoreaverage score · variant=Codingofficial-benchmark57
LiveBench Data Analysisgpt-6-astraOpenAI82.97average scoreaverage score · variant=Data Analysisofficial-benchmark57
LiveBench Instruction FollowingGemini 3.8 FlashGoogle81.41average scoreaverage score · variant=IFofficial-benchmark57
LiveBench LanguageClaude Fable 5Anthropic90.68average scoreaverage score · variant=Languageofficial-benchmark57
LiveBench MathematicsClaude Fable 5.1Anthropic97.01average scoreaverage score · variant=Mathematicsofficial-benchmark57
LiveBench Reasoninggpt-6-astraOpenAI92.65average scoreaverage score · variant=Reasoningofficial-benchmark57
MMMU-Progpt-6-astraOpenAI86.88accuracyaccuracy · evaluator=Artificial Analysisindependent-evaluator156
SWE-bench (full test split)claude-3-opusAnthropic3.79resolvedresolved · board=Test · system=RAG baselineofficial-benchmark6
SWE-bench Liteclaude-3-opusAnthropic4.33resolvedresolved · board=Lite · system=RAG baselineofficial-benchmark6
SWE-bench Multilingualgemini-3-flashGoogle72.7resolvedresolved · board=Multilingual · system=mini-SWE-agentofficial-benchmark13
SWE-bench Multimodalo3OpenAI35.98resolvedresolved · board=Multimodal · system=GUIRepairofficial-benchmark4
SWE-bench VerifiedClaude Opus 4.5Anthropic76.8resolvedresolved · board=Verified · system=mini-SWE-agentofficial-benchmark42
SciCodeClaude Fable 5.1Anthropic63.08accuracyaccuracy · evaluator=Artificial Analysisindependent-evaluator126
Terminal-Benchgpt-5.6-solOpenAI65.91accuracyaccuracy · variant=hard · evaluator=Artificial Analysisindependent-evaluator315
τ²-benchZ.ai GLM 5.2Z.ai (Zhipu AI)99.12pass^1pass^1 · evaluator=Artificial Analysisindependent-evaluator323

One row per benchmark: the best current row of the primary comparability group among results observed by the date. See the benchmark's Frontier tab for the full leader history.

leaders from results observed by the date (evaluation dates are not used: a result is known only once observed)

Hardware

Hardware as of 11 Sept 2026 25

All →
Hardware as of 2026-09-11
HardwareManufacturerKindMemoryBandwidthReleasedBasis
Apple M3 UltraApplesoc96 / 256 / 512 GB819 GB/sMar 2025observed
NVIDIA GeForce RTX 5090NVIDIAgpu32 GB1,792 GB/sJan 2025observed
NVIDIA DGX SparkNVIDIAsystem128 GB273 GB/s2025observed
Apple M4 ProApplesoc24 / 48 / 64 GB273 GB/sOct 2024observed
Apple M4 MaxApplesoc36 / 48 / 64 / 128 GB546 GB/sOct 2024observed
AMD Instinct MI325XAMDgpu256 GB6,000 GB/sOct 2024observed
Apple M4Applesoc16 / 24 / 32 GB120 GB/sMay 2024observed
NVIDIA B200NVIDIAgpu180 GB—Mar 2024observed
NVIDIA H200NVIDIAgpu141 GB4,800 GB/s2024observed
AMD Instinct MI300XAMDgpu192 GB5,300 GB/sDec 2023observed
Apple M3Applesoc8 / 16 / 24 GB100 GB/sOct 2023observed
Apple M3 ProApplesoc18 / 36 GB150 GB/sOct 2023observed
Apple M3 MaxApplesoc36 / 48 / 64 / 96 / 128 GB400 GB/sOct 2023observed
Apple M2 UltraApplesoc64 / 128 / 192 GB800 GB/sJun 2023observed
Apple M2 ProApplesoc16 / 32 GB200 GB/sJan 2023observed
Apple M2 MaxApplesoc32 / 64 / 96 GB400 GB/sJan 2023observed
NVIDIA GeForce RTX 4090NVIDIAgpu24 GB1,008 GB/sOct 2022observed
NVIDIA H100 SXMNVIDIAgpu80 GB3,350 GB/sSept 2022observed
Apple M2Applesoc8 / 16 / 24 GB100 GB/sJun 2022observed
Apple M1 UltraApplesoc64 / 128 GB800 GB/sMar 2022observed
Apple M1 ProApplesoc16 / 32 GB200 GB/sOct 2021observed
Apple M1 MaxApplesoc32 / 64 GB400 GB/sOct 2021observed
Apple M1Applesoc8 / 16 GB68 GB/sNov 2020observed
NVIDIA A100 80GBNVIDIAgpu80 GB2,039 GB/sNov 2020observed
NVIDIA GeForce RTX 3090NVIDIAgpu24 GB936 GB/sSept 2020observed

Shareable: this URL reproduces the view. Per-entity: every entity page has a History tab with the same as-of reconstruction. What changed since? Diff 11 Sept 2026 → today. Methodology →