https://www.vaanalytics.in Every AI model. Every provider. Every change. Changes to AI model prices, capabilities and availability, plus daily model news. Mon, 14 Sep 2026 02:47:01 GMT https://www.vaanalytics.in/m/deepseek/deepseek-v4.1-flash 1eb270cbe0c7 Mon, 14 Sep 2026 00:00:00 GMT cost.input: 0.15 → 0.3; cost.output: 0.6 → 1.2; cost.cache_read: 0.003 → 0.006 https://www.vaanalytics.in/changelog 35349711dff4 Mon, 14 Sep 2026 00:00:00 GMT cost.input: 1.122 → 0.5808; cost.output: 3.366 → 1.7424 https://www.vaanalytics.in/changelog 51b631a1b00e Mon, 14 Sep 2026 00:00:00 GMT limit.context: 202752 → 1048576; limit.output: 182476 → 131072 https://www.vaanalytics.in/changelog 5fa835341737 Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.23 https://www.vaanalytics.in/changelog 65d5d84e0363 Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.05 https://www.vaanalytics.in/changelog 7bc49ea7f58f Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.23 https://www.vaanalytics.in/m/nvidia/nemotron-3-ultra-550b-a55b 928be812a7d0 Mon, 14 Sep 2026 00:00:00 GMT limit.context: 256000 → 202800; limit.output: 32768 → 182520 https://www.vaanalytics.in/changelog 932208afb6e2 Mon, 14 Sep 2026 00:00:00 GMT limit.output: 131072 → 943717; cost.input: 1.092 → 1.4; cost.output: 3.432 → 4.4; cost.cache_read: 0.2028 → 0.26 https://www.vaanalytics.in/m/deepseek/deepseek-v4-pro-0813 9a6f5bf38494 Mon, 14 Sep 2026 00:00:00 GMT limit.context: 1048576 → 1024000; limit.output: 393216 → 384000 https://www.vaanalytics.in/changelog 9ca353232554 Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.03 https://www.vaanalytics.in/changelog 9ec39cc50151 Mon, 14 Sep 2026 00:00:00 GMT cost.input: 0.352 → 0.176; cost.output: 1.056 → 0.528 https://www.vaanalytics.in/changelog a66be6094429 Mon, 14 Sep 2026 00:00:00 GMT limit.output: 8192 → 16384; cost.input: 0.2275 → 0.12; cost.output: 0.91 → 0.24 https://www.vaanalytics.in/m/deepseek/deepseek-v4-flash-0731 a91dc4f5dc7a Mon, 14 Sep 2026 00:00:00 GMT cost.input: 0.04 → 0.06; cost.output: 0.08 → 0.12; cost.cache_read: 0.008 → 0.012 https://www.vaanalytics.in/changelog ab4198e65feb Mon, 14 Sep 2026 00:00:00 GMT limit.context: 131072 → 40960; limit.output: 8192 → 16384 https://www.vaanalytics.in/m/deepseek/deepseek-v4-pro-0813 ac1c0141e4d4 Mon, 14 Sep 2026 00:00:00 GMT limit.output: 393216 → 384000; cost.input: 0.57816 → 1.05468; cost.output: 1.73448 → 3.16404; cost.cache_read: 0.018396 → 0.035156 https://www.vaanalytics.in/changelog af4da3899529 Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.05 https://www.vaanalytics.in/m/nvidia/nemotron-3-ultra-550b-a55b b0dd435dc05d Mon, 14 Sep 2026 00:00:00 GMT limit.output: 32768 → 182520; cost.input: 0.625 → 0.6; cost.output: 3.125 → 2.4; cost.cache_read: 0.1875 → 0.12 https://www.vaanalytics.in/changelog b7880a364b0a Mon, 14 Sep 2026 00:00:00 GMT cost.cache_read: null → 0.03 https://www.vaanalytics.in/m/deepseek/deepseek-v4-flash dc7a20bd872f Mon, 14 Sep 2026 00:00:00 GMT cost.input: 0.049 → 0.089866; cost.output: 0.098 → 0.179732; cost.cache_read: 0.0098 → 0.017973 https://www.vaanalytics.in/changelog e6588348c335 Mon, 14 Sep 2026 00:00:00 GMT model_added: DeepSeek V4.1 Flash (Consensus Protocol) via llmgateway-providers https://www.vaanalytics.in/changelog f662869dc91c Mon, 14 Sep 2026 00:00:00 GMT limit.output: 182476 → 131072; cost.input: 0.6 → 0.6832; cost.output: 2 → 2.1472; cost.cache_read: 0.15 → 0.12688 https://www.vaanalytics.in/changelog fe0148ee67bd Mon, 14 Sep 2026 00:00:00 GMT limit.context: 1048576 → 1048575; limit.output: 131072 → 943717 https://shattered.io/abacus-ai-smaug-open-weight-models-2026 1pg2pai Sun, 13 Sep 2026 00:45:43 GMT Smaug Mini is the smallest of the three, based on Qwen3.8 27B and aimed at compact multimodal tasks where a lighter model is cheaper to run at scale. Abacus.AI describes Smaug Mini as suited to smaller, high-volume jobs rather than the deep multi-step reasoning chains that Smaug Agentic is built for. Between the three, Abacus.AI is effectively offering a size ladder: pick Mini for cheap, high-volume multimodal tasks, Flash for always-on agents with long context, and Agentic for the hardest [...] According to Abacus.AI’s own materials on its open-source page, the underlying methodology combines human-curated, real-world agentic traces with synthetic data grounded in hard examples, then applies that single training recipe across three different open-weight bases: Smaug Flash on DeepSeek V4 Flash, Smaug Mini on Qwen3.8 27B, and Smaug Agentic on Kimi K3. That’s a notable design choice: rather than building one model and shrinking it, Abacus.AI runs the same fine-tuning process across three [...] | Model | Base model | Parameter scale | Primary use case | --- --- | | Smaug Agentic | Kimi K3 | 2 trillion | Complex coding, long-running agentic loops | | Smaug Flash | DeepSeek V4 Flash | Not disclosed | Always-on enterprise agents, long context, heavy tool use | | Smaug Mini | Qwen3.8 27B | 27 billion | Compact multimodal tasks, high-volume jobs | ## The Fine-Tuning Technique Behind the 15-20% Gain https://www.vaanalytics.in/changelog 002030f6509a Sun, 13 Sep 2026 00:00:00 GMT last_updated: "2025-12-02" → "2025-12-01" https://www.vaanalytics.in/m/google/gemma-3-12b-it 012ca1bb6938 Sun, 13 Sep 2026 00:00:00 GMT name: "Gemma 3 12B" → "Gemma 3 12B IT"; knowledge: "2024-08-31" → "2024-08"; release_date: "2025-03-13" → "2025-03-12"; last_updated: "2025-03-13" → "2025-03-12" https://www.vaanalytics.in/changelog 0142b03cf108 Sun, 13 Sep 2026 00:00:00 GMT cost.input: 0.3 → 0.2; cost.output: 1.2 → 0.6 https://www.vaanalytics.in/m/google/gemma-3-4b-it 0148d3a8a93d Sun, 13 Sep 2026 00:00:00 GMT model_added: Gemma 3 4B via merge-gateway https://www.vaanalytics.in/changelog 0184e8db3d25 Sun, 13 Sep 2026 00:00:00 GMT cost.output: 2 → 2.2; cost.cache_read: 0.08 → 0.11 https://www.vaanalytics.in/m/deepseek/deepseek-v4-flash-0423 01a5b0eae2f7 Sun, 13 Sep 2026 00:00:00 GMT cost.input: 0.44 → 0.19; cost.output: 1.32 → 0.51; cost.cache_read: 0.014 → 0.028 https://www.vaanalytics.in/changelog 04715ac8f9ce Sun, 13 Sep 2026 00:00:00 GMT name: "Qwen3 Coder 30B A3B Instruct" → "Qwen3-Coder 30B-A3B Instruct"; description: "Qwen coding model for software agents, repository edits, and code reasoning" → "Smaller Qwen coder for efficient local agents and repo-level fixes"; open_weights: false → true; knowledge: "2024-04" → "2025-04"; release_date: "2025-09-18" → "2025-07-31" https://www.vaanalytics.in/changelog 04c694766e33 Sun, 13 Sep 2026 00:00:00 GMT model_added: Kimi K3 (Tencent Cloud) via llmgateway-providers https://www.vaanalytics.in/changelog 050cdcf27a8b Sun, 13 Sep 2026 00:00:00 GMT model_added: Gemma 3 27B IT (Amazon Bedrock) via edenai https://www.vaanalytics.in/changelog 0666a2c7a3d8 Sun, 13 Sep 2026 00:00:00 GMT tool_call: false → true https://www.vaanalytics.in/changelog 0758a77cdb62 Sun, 13 Sep 2026 00:00:00 GMT attachment: true → false https://www.vaanalytics.in/m/mistral/mistral-large-2512 081cd5970528 Sun, 13 Sep 2026 00:00:00 GMT release_date: "2024-11-01" → "2025-12-02" https://www.vaanalytics.in/changelog 0842a5611c53 Sun, 13 Sep 2026 00:00:00 GMT description: "Compact Mistral model for edge, latency-sensitive, and cost-efficient workloads" → "Open vision-language model for efficient local deployment, instruction following, and tool use"; attachment: false → true; open_weights: false → true; release_date: "2024-12-01" → "2025-12-02"; last_updated: "2024-12-01" → "2025-12-02"; modalities.input: "text" → "text,image" https://www.vaanalytics.in/changelog 0939a6655a25 Sun, 13 Sep 2026 00:00:00 GMT model_added: GLM-5V-Turbo via llmgateway https://www.vaanalytics.in/changelog 0949f26c56c4 Sun, 13 Sep 2026 00:00:00 GMT description: "Open multimodal Llama model for long-context analysis and efficient agents" → "Open Llama with long-context vision for efficient multimodal agents"; limit.context: 3500000 → 10000000; limit.output: 16384 → 8192 https://www.vaanalytics.in/changelog 0ae3364e5460 Sun, 13 Sep 2026 00:00:00 GMT model_added: GLM-5.3 via agentrouter https://www.vaanalytics.in/changelog 0b124d663437 Sun, 13 Sep 2026 00:00:00 GMT model_added: Voxtral Mini 3B 2507 (Amazon Bedrock, US) via edenai https://www.vaanalytics.in/m/google/gemma-3-4b-it 0ba5d037fca0 Sun, 13 Sep 2026 00:00:00 GMT model_added: Gemma 3 4B IT via deepinfra https://www.vaanalytics.in/changelog 0bcc4f2e3763 Sun, 13 Sep 2026 00:00:00 GMT cost.cache_read: 3 → null; cost.tiers: [{"minContext":272000,"rates":{"input":60,"output":270,"cacheRead":6,"cacheWrite":null,"reasoning":null,"inputAudio":null,"outputAudio":null}}] → [{"minContext":272000,"rates":{"input":60,"output":270,"cacheRead":null,"cacheWrite":null,"reasoning":null,"inputAudio":null,"outputAudio":null}}] https://www.vaanalytics.in/changelog 0bd7fc440b7a Sun, 13 Sep 2026 00:00:00 GMT model_added: inclusionAI: Ling 3.0 Flash VL via kilo https://www.vaanalytics.in/changelog 0c7db6916b2f Sun, 13 Sep 2026 00:00:00 GMT name: "Qwen/Qwen3-VL-235B-A22B-Instruct" → "Qwen3 VL 235B A22B Instruct"; description: "Qwen vision-language model for visual reasoning, documents, and agent tasks" → "Qwen vision-language instruct model for visual reasoning, documents, and agent tasks"; open_weights: false → true; knowledge: null → "2025-03-31"; release_date: "2025-10-04" → "2025-09-23"; limit.context: 262000 → 262144; cost.input: 0.3 → 0.53; cost.output: 1.5 → 2.66 https://www.vaanalytics.in/changelog 0c8c56401ccf Sun, 13 Sep 2026 00:00:00 GMT reasoning: false → true https://www.vaanalytics.in/changelog 0c92d446aa29 Sun, 13 Sep 2026 00:00:00 GMT attachment: false → true; modalities.input: "text" → "text,image"; limit.context: 131072 → 256000; limit.output: 32768 → 64000 https://www.vaanalytics.in/changelog 0dde0cc6224a Sun, 13 Sep 2026 00:00:00 GMT model_added: DeepSeek V4 Flash (Tencent Cloud) via llmgateway-providers https://www.vaanalytics.in/changelog 0f0c29286b69 Sun, 13 Sep 2026 00:00:00 GMT description: "Open-weight GPT model for self-hosted reasoning and instruction-following workloads" → "Open GPT reasoning model for self-hosted agents and controllable deployments" https://www.vaanalytics.in/changelog 0f83f265f75a Sun, 13 Sep 2026 00:00:00 GMT limit.context: 3500000 → 10000000; limit.output: 16384 → 8192 https://www.vaanalytics.in/changelog 0f849e7fd071 Sun, 13 Sep 2026 00:00:00 GMT model_added: Hy-MT2 Plus via llmgateway