{"slug":"deepseek-pro","label":"DeepSeek Pro","item_count":3,"day_count":3,"source_count":2,"first_seen":"2026-07-19T01:41:33+00:00","last_updated":"2026-07-31T18:27:25+00:00","generated_at":"2026-07-31T20:05:13.706734+00:00","sources":["latent_space","search_cn_open_weight_labs"],"days":[{"date":"2026-07-19","items":[{"title":"Kimi K3 vs DeepSeek V4 Pro vs GLM-5.2: Open Trillion-Scale MoE Models Compared on Benchmarks, License, and Serving Cost - MarkTechPost","url":"https://news.google.com/rss/articles/CBMi7gFBVV95cUxNd2hDN3VQRUVsNS1YcjhtbGF5azdBTXBXSk9JWnRwbFNQWmVlbExuZ3BZMFg2Z3lWMlhxdWJ1U1BPcjFoRGpPajQxVDZic2U5TWY2RVhkYW83SDJJNWx5UjVwVkdreThodEt4SXpwamlNSlJ2V0dwOU9CblpIN3dyZHJGVXVPaldDVHo5bTRodVI1RzlfYzYyNlVENXZ5Z3FfVXNKQzhabHRaOHZZSFlqQ295QXNOS28wZVBieU5yRDRic01YaTRJUmhQSncxazJUbGx5NHlXbjdPUzNYNmR5VF9jTVJTMkk1SXgtWnBB?oc=5","source":"search_cn_open_weight_labs","type":"news","summary_1line":"Kimi K3 vs DeepSeek V4 Pro vs GLM-5.2: Open Trillion-Scale MoE Models Compared on Benchmarks, License, and Serving Cost MarkTechPost","sid":"963336aecd23c1c4","published":"2026-07-19T01:41:33+00:00","editor_note":"First independent three-way comparison of V4 Pro against Kimi K3 and GLM-5.2 on benchmarks, licensing, and serving cost."}]},{"date":"2026-07-23","items":[{"title":"[AINews] \"Laguna S 2.1 Released: Cheaper than Deepseek v4 Flash, Better than V4 Pro\"","url":"https://www.latent.space/p/ainews-laguna-s-21-released-cheaper","source":"latent_space","type":"release","summary_1line":"a quiet day lets us highlight a new neolab win.","sid":"b84094e090b7831f","published":"2026-07-23T05:18:47+00:00","editor_note":"A rival low-cost model surfaces already beating V4 Pro on price-performance."}]},{"date":"2026-07-31","items":[{"title":"DeepSeek Retrained V4-Flash Beats Its Flagship Pro on Nine Agent Benchmarks - Tech Times","url":"https://news.google.com/rss/articles/CBMixgFBVV95cUxPeFMyNVd0Y2J3MDJoMTc5MEw3ajZTNDBjSmszd0xaNktucU5WTTdyYmpuY2R2UjhfTEx6MjdhWHhwa0l6NU02VDFfdmMySktYYmx0eUZZU3FrczJ0OGlTbWxHNHdkY04weTQtRExSM1dmWlgxaUpfMTc2UHp6NFdNOTRrRkVUenFlVzFSbWZsVm9vbHR2SHRzSkN3dXh1aHVyN0dMSlZNa1NHdWYyS2RLMXRGR3RvX3M0TGppeWZUc2daZ09UMWc?oc=5","source":"search_cn_open_weight_labs","type":"news","summary_1line":"DeepSeek Retrained V4-Flash Beats Its Flagship Pro on Nine Agent Benchmarks Tech Times","why_it_matters":"Matches feed focus: agent.","sid":"2281120b69fc276a","published":"2026-07-31T18:27:25+00:00","editor_note":"DeepSeek's retrained V4-Flash overtakes flagship V4 Pro on nine agent benchmarks."}]}],"editorial":{"tldr":"DeepSeek's flagship V4 Pro entered late July already under scrutiny, benchmarked head-to-head against Kimi K3 and GLM-5.2 on cost, license, and serving price. Within days a rival low-cost model was already beating it on price-performance.","stale":false,"whats_new":"DeepSeek's own retrained V4-Flash — the cheaper, faster sibling model — now beats flagship V4 Pro on nine agent benchmarks.","why_it_matters":"Teams running DeepSeek for agent workloads should re-benchmark before assuming Pro is still the right default: the higher-priced flagship no longer leads on agent tasks against its own cheaper sibling.","take_for_builders":"Running DeepSeek V4 Pro for agent workloads? Re-benchmark against V4-Flash before your next deployment — DeepSeek's own numbers now put the cheaper sibling ahead on agent tasks.","status":{"state":"Developing","tone":"rising","changed":"2026-07-31","detail":"DeepSeek's cheaper V4-Flash variant has overtaken flagship V4 Pro on agent benchmarks, following earlier comparisons that were already unfavorable to Pro on price-performance.","track":[{"label":"compared","detail":"Jul 19","tone":"neutral","weight":33},{"label":"undercut","detail":"Jul 23","tone":"rising","weight":33},{"label":"overtaken","detail":"Jul 31 → now","tone":"turn","weight":34}]},"beats":[{"kicker":"COMPARISON","tone":"neutral","headline":"V4 Pro benchmarked against Kimi K3 and GLM-5.2 on cost, license, and performance","summary":"A three-way trillion-scale MoE comparison put V4 Pro alongside two open-weight rivals on benchmarks, licensing terms, and serving cost.","sids":["963336aecd23c1c4"]},{"kicker":"PRESSURE","tone":"rising","headline":"Rival Laguna S 2.1 undercuts V4 Pro on price and performance","summary":"A newly released low-cost model was pitched as cheaper than DeepSeek's own V4-Flash and better than V4 Pro.","sids":["b84094e090b7831f"]},{"kicker":"THE TURN","tone":"turn","headline":"DeepSeek's retrained V4-Flash beats V4 Pro on nine agent benchmarks","summary":"DeepSeek's cheaper sibling model now outperforms the flagship Pro on agent-focused evaluations.","sids":["2281120b69fc276a"]}],"open_questions":["Does DeepSeek reposition V4-Flash as the default choice for agent workloads, or keep Pro as the flagship despite the benchmark loss?","Do independent evals confirm V4-Flash's nine-benchmark lead over Pro, or is this DeepSeek's own reported result?","Does V4-Flash's pricing undercut V4 Pro enough to change the serving-cost calculus for teams already running Pro?"],"generated_at":"2026-07-31T20:04:34+00:00"}}