The NNPDF Collaboration presents the first determination of nucleon parton distributions at NLO and NNLO based on a global data set including LHC data: NNPDF2.3. The data set includes deep inelastic, Drell-Yan, gauge boson production, and jet data, as well as all relevant LHC data with available systematic uncertainties. An improved FastKernel method is introduced for fitting to this extended data set and for more effective minimization. The NNPDF2.3 PDF sets are compared to NNPDF2.1 to assess the impact of LHC data. All LHC data are broadly consistent with each other and with older data sets. Predictions for standard cross-sections are presented, with a focus on the impact of ATLAS electroweak data on the proton's strangeness fraction. Collider PDF sets, constructed from HERA, Tevatron, and LHC data, are also presented, though they are not precise or complete enough for competitive PDF determination. The paper discusses methodological improvements, results, and the impact of LHC data on parton distributions. It also presents statistical features, parton distributions, and comparisons with previous sets. The analysis shows that LHC data significantly improve the accuracy of PDF determinations, though collider-only data sets are less precise. The NNPDF2.3 PDFs are provided for various values of αs, with improved methodology and data inclusion. The results demonstrate the effectiveness of the new method in accurately determining parton distributions.The NNPDF Collaboration presents the first determination of nucleon parton distributions at NLO and NNLO based on a global data set including LHC data: NNPDF2.3. The data set includes deep inelastic, Drell-Yan, gauge boson production, and jet data, as well as all relevant LHC data with available systematic uncertainties. An improved FastKernel method is introduced for fitting to this extended data set and for more effective minimization. The NNPDF2.3 PDF sets are compared to NNPDF2.1 to assess the impact of LHC data. All LHC data are broadly consistent with each other and with older data sets. Predictions for standard cross-sections are presented, with a focus on the impact of ATLAS electroweak data on the proton's strangeness fraction. Collider PDF sets, constructed from HERA, Tevatron, and LHC data, are also presented, though they are not precise or complete enough for competitive PDF determination. The paper discusses methodological improvements, results, and the impact of LHC data on parton distributions. It also presents statistical features, parton distributions, and comparisons with previous sets. The analysis shows that LHC data significantly improve the accuracy of PDF determinations, though collider-only data sets are less precise. The NNPDF2.3 PDFs are provided for various values of αs, with improved methodology and data inclusion. The results demonstrate the effectiveness of the new method in accurately determining parton distributions.