1324return new TiktokenTokenizer( 1352=> new TiktokenTokenizer(vocabFilePath, preTokenizer, specialTokens, normalizer, cacheSize); 1372=> new TiktokenTokenizer(vocabStream, preTokenizer, specialTokens, normalizer, cacheSize); 1403return new TiktokenTokenizer(encoder, decoder, vocab, preTokenizer, specialTokens, normalizer, cacheSize); 1467return new TiktokenTokenizer(vocabStream,

21private static readonly Tokenizer _gpt4oTokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

12private static readonly Tokenizer _tokenizer = TiktokenTokenizer.CreateForModel("gpt-4");

39HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4"))); 71HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 13 }); 96HeaderChunker lessThanContext = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 5 }); 99HeaderChunker sameAsContext = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 6 }); 119HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 30 }); 136HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 37 }); 146HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 100 }); 175Tokenizer tokenizer = TiktokenTokenizer.CreateForModel("gpt-4");

16var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

17var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o"); 25var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

16var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

26Tokenizer tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

226private static IngestionChunker<string> CreateChunker() => new HeaderChunker(new(TiktokenTokenizer.CreateForModel("gpt-4")));

43/// Create <see cref="TiktokenTokenizer"/> from tokenizer model file. 47public static TiktokenTokenizer FromPretrained( 53return TiktokenTokenizer.Create(File.OpenRead(modelFilePath), preTokenizer, normalizer: null, specialTokens: _specialTokens);

68var tokenizer = LlamaTokenizerHelper.FromPretrained(modelWeightFolder);

36var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 39var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

81public static ICausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM> LoadModel(string weightFolder, string checkPointName = "model.safetensors.index.json") 91var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 94var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

37var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 40var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

1879new RegexPreTokenizer(TiktokenTokenizer.P50kBaseRegex(), CodeGenTokenizer.CodeGenSpecialTokens),

117vocabStream, mergesStream, new RegexPreTokenizer(TiktokenTokenizer.P50kBaseRegex(), CodeGenTokenizer.CodeGenSpecialTokens), normalizer: null,

1291private static TiktokenTokenizer CreateForModel( 1346public static TiktokenTokenizer Create( 1366public static TiktokenTokenizer Create( 1387public static async Task<TiktokenTokenizer> CreateAsync( 1419public static async Task<TiktokenTokenizer> CreateAsync( 1445public static TiktokenTokenizer CreateForModel( 1484public static async Task<TiktokenTokenizer> CreateForModelAsync( 1521public static TiktokenTokenizer CreateForModel(string modelName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1531public static TiktokenTokenizer CreateForEncoding(string encodingName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1565throw new ArgumentException($"The encoding name '{encodingName}' is not supported. The only supported encoding names are: {TiktokenTokenizer.Cl100kBaseEncodingName}, {TiktokenTokenizer.P50kBaseEncodingName}, {TiktokenTokenizer.P50kEditEncodingName}, and {TiktokenTokenizer.R50kBaseEncodingName}.", nameof(encodingName));

32return SplitText(text, TiktokenTokenizer.P50kBaseRegex()); 47return SplitText(text, TiktokenTokenizer.P50kBaseRegex());

30var exception = Record.Exception(() => TiktokenTokenizer.CreateForModel(modelName)); 53TiktokenTokenizer externalTokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, preTokenizer: null, normalizer: null);

32public static Tokenizer GPT4 { get; } = TiktokenTokenizer.CreateForModel("gpt-4", _specialTokens); 33public static Tokenizer GPT2 { get; } = TiktokenTokenizer.CreateForModel("gpt2"); 34public static Tokenizer P50kBase { get; } = TiktokenTokenizer.CreateForModel("text-davinci-003"); 35public static Tokenizer R50kBase { get; } = TiktokenTokenizer.CreateForModel("ada"); 36public static Tokenizer P50kEdit { get; } = TiktokenTokenizer.CreateForModel("text-davinci-edit-001"); 37public static Tokenizer GPT4o { get; } = TiktokenTokenizer.CreateForModel("gpt-4o"); 38public static Tokenizer GPT5 { get; } = TiktokenTokenizer.CreateForModel("gpt-5"); 39public static Tokenizer GPT5_1 { get; } = TiktokenTokenizer.CreateForModel("gpt-5.1"); 40public static Tokenizer GPT5_2 { get; } = TiktokenTokenizer.CreateForModel("gpt-5.2"); 41public static Tokenizer GPT5_3 { get; } = TiktokenTokenizer.CreateForModel("gpt-5.3"); 42public static Tokenizer Phi4 { get; } = TiktokenTokenizer.CreateForModel("phi-4"); 43public static TiktokenTokenizer GptOss { get; } = TiktokenTokenizer.CreateForModel("gpt-oss-20b"); 51Assert.True(GPT4 is TiktokenTokenizer); 52IReadOnlyDictionary<string, int>? specialTokens = (GPT4 as TiktokenTokenizer)!.SpecialTokens; 56string assemblyName = typeof(TiktokenTokenizer).Assembly.FullName!; 67Tokenizer tokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, GPT4.PreTokenizer, null, specialTokens); 72tokenizer = TiktokenTokenizer.Create(stream, GPT4.PreTokenizer, null, specialTokens); 76tokenizer = await TiktokenTokenizer.CreateAsync(tokenizerDataFileName, GPT4.PreTokenizer, normalizer: null, specialTokens); 81tokenizer = await TiktokenTokenizer.CreateAsync(stream, GPT4.PreTokenizer, normalizer: null, specialTokens); 87tokenizer = TiktokenTokenizer.CreateForModel("gpt-4", stream); 93tokenizer = await TiktokenTokenizer.CreateForModelAsync("gpt-3.5-turbo", stream); 97tokenizer = TiktokenTokenizer.CreateForModel("gpt-4"); 124TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 125TiktokenTokenizer externalTokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, tokenizer.PreTokenizer, null, tiktoken.SpecialTokens); 149TestDecodingWithSpan((tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 166private void TestDecodingWithSpan(TiktokenTokenizer tokenizer, int[] ids, string expectedDecoded) 202TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 233TestDecodingWithSpan((gpt4Tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 245TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 280TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 292foreach (TiktokenTokenizer tokenizer in new[] { GPT4o, GptOss, GPT5, GPT5_1, GPT5_2 }) 345TestDecodingWithSpan((GPT2 as TiktokenTokenizer)!, encoded.ToArray(), text); 364TestDecodingWithSpan((P50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 383TestDecodingWithSpan((P50kEdit as TiktokenTokenizer)!, encoded.ToArray(), text); 402TestDecodingWithSpan((R50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 486Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(modelName); 487Assert.True(tokenizer is TiktokenTokenizer); 500Tokenizer tokenizer = TiktokenTokenizer.CreateForEncoding(encodingName); 501Assert.True(tokenizer is TiktokenTokenizer); 515Tokenizer tokenizer1 = TiktokenTokenizer.CreateForModel(modelName); 517Assert.True(tokenizer is TiktokenTokenizer); 518Assert.True(tokenizer1 is TiktokenTokenizer); 520TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 521TiktokenTokenizer tiktoken1 = (tokenizer1 as TiktokenTokenizer)!; 532Assert.Throws<ArgumentNullException>(() => TiktokenTokenizer.CreateForEncoding(null!)); 533Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("r50k_base_")); 534Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_base_")); 535Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_edit_")); 536Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("cl100k_base_")); 537Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("o200k_base_")); 538Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("o200k_harmony_")); 565Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(name); 566Assert.True(tokenizer is TiktokenTokenizer); 570int entriesCount = GetEncoder((tokenizer as TiktokenTokenizer)!)!.Count; 847private static IReadOnlyDictionary<ReadOnlyMemory<byte>, int>? GetEncoder(TiktokenTokenizer tiktoken) 848=> typeof(TiktokenTokenizer).GetProperty("Encoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<ReadOnlyMemory<byte>, int>; 850private static IReadOnlyDictionary<int, ReadOnlyMemory<byte>>? GetDecoder(TiktokenTokenizer tiktoken) 851=> typeof(TiktokenTokenizer).GetProperty("Decoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<int, ReadOnlyMemory<byte>>; 853private static IReadOnlyDictionary<string, int>? GetVocabulary(TiktokenTokenizer tiktoken) 854=> typeof(TiktokenTokenizer).GetProperty("Vocabulary", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<string, int>;